You will be redirected to the main page within 3 seconds. If not redirected, please click here.
+ Page not found | HKU NLP Group
Page not found
Looks like there has been a mistake. Nothing exists here.
You will be redirected to the main page within 3 seconds. If not redirected, please click here.
\ No newline at end of file
diff --git a/Gemfile b/Gemfile
deleted file mode 100644
index a8764ec..0000000
--- a/Gemfile
+++ /dev/null
@@ -1,25 +0,0 @@
-source 'https://rubygems.org'
-group :jekyll_plugins do
- gem 'jekyll'
- gem 'jekyll-archives'
- gem 'jekyll-diagrams'
- gem 'jekyll-email-protect'
- gem 'jekyll-feed'
- gem 'jekyll-imagemagick'
- gem 'jekyll-minifier'
- gem 'jekyll-paginate-v2'
- gem 'jekyll-scholar'
- gem 'jekyll-sitemap'
- gem 'jekyll-target-blank'
- gem 'jekyll-twitter-plugin'
- gem 'jekyll-redirect-from'
- gem 'jemoji'
- gem 'mini_racer'
- gem 'unicode_utils'
- gem 'webrick'
- gem 'jekyll-loading-lazy'
-end
-group :other_plugins do
- gem 'httparty'
- gem 'feedjira'
-end
diff --git a/_bibliography/papers.bib b/_bibliography/papers.bib
deleted file mode 100644
index 10db8b7..0000000
--- a/_bibliography/papers.bib
+++ /dev/null
@@ -1,580 +0,0 @@
----
----
-
-@article{Xie2023OpenAgentsAO,
- title = {OpenAgents: An Open Platform for Language Agents in the Wild},
- author = {Tianbao Xie and Fan Zhou and Zhoujun Cheng and Peng Shi and Luoxuan Weng and Yitao Liu and Toh Jing Hua and Junning Zhao and Qian Liu and Che Liu and Leo Z. Liu and Yiheng Xu and Hongjin Su and Dongchan Shin and Caiming Xiong and Tao Yu},
- journal = {Conference on Language Modeling (COLM)},
- year = {2024},
- arxiv = {2310.10634},
- code = {https://github.com/xlang-ai/OpenAgents},
- preview = {Xie2023OpenAgentsAO.png}
-}
-
-@article{Sun2023CorexPT,
- title = {Corex: Pushing the Boundaries of Complex Reasoning through Multi-Model Collaboration},
- author = {Qiushi Sun and Zhangyue Yin and Xiang Li and Zhiyong Wu and Xipeng Qiu and Lingpeng Kong},
- journal = {Conference on Language Modeling (COLM)},
- year = {2024},
- arxiv = {2310.00280},
- code = {https://github.com/QiushiSun/Corex},
- preview = {Sun2023CorexPT.png}
-}
-
-
-@article{Gao2024CoCA,
- title = {CoCA: Regaining Safety-awareness of Multimodal Large Language Models with Constitutional Calibration},
- author = {Jiahui Gao and Renjie Pi and Tianyang Han and Han Wu and Lanqing HONG and Lingpeng Kong and Xin Jiang and Zhenguo Li},
- journal = {Conference on Language Modeling (COLM)},
- year = {2024}
-}
-
-
-@article{Zhao2024EmpoweringLL,
- title = {Empowering Large Language Model Agents through Action Learning},
- author = {Haiteng Zhao and Chang Ma and Guoyin Wang and Jing Su and Lingpeng Kong and Jingjing Xu and Zhi-Hong Deng and Hongxia Yang},
- journal = {Conference on Language Modeling (COLM)},
- year = {2024},
- arxiv = {2402.15809},
- code = {https://github.com/zhao-ht/LearnAct},
- preview = {Zhao2024EmpoweringLL.png}
-}
-
-@article{zheng2023reparameterized,
- title = {A Reparameterized Discrete Diffusion Model for Text Generation},
- author = {Lin Zheng and Jianbo Yuan and Lei Yu and Lingpeng Kong},
- journal = {Conference on Language Modeling (COLM)},
- year = {2024},
- arxiv = {2302.04542},
- code = {https://github.com/HKUNLP/reparam-discrete-diffusion},
- preview = {zheng2023reparameterized.png}
-}
-
-
-@article{Zhao2023SEGOSS,
- title = {SEGO: Sequential Subgoal Optimization for Mathematical Problem-Solving},
- author = {Xueliang Zhao and Xinting Huang and Wei Bi and Lingpeng Kong},
- journal = {In Proceedings of the
- Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2024},
- arxiv = {2310.12960},
- code = {https://github.com/zhaoxlpku/SEGO},
- preview = {Zhao2023SEGOSS.png}
-}
-
-@article{Zhao2024BBABB,
- title = {BBA: Bi-Modal Behavioral Alignment for Reasoning with Large Vision-Language Models},
- author = {Xueliang Zhao and Xinting Huang and Tingchen Fu and Qintong Li and Shansan Gong and Lemao Liu and Wei Bi and Lingpeng Kong},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2024},
- arxiv = {2402.13577},
- preview = {Zhao2024BBABB.png}
-}
-
-@article{Li2024RedTV,
- title = {Red Teaming Visual Language Models},
- author = {Mukai Li and Lei Li and Yuwei Yin and Masood Ahmed and Zhenguang Liu and Qi Liu},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2024},
- arxiv = {2401.12915},
- code = {https://huggingface.co/datasets/MMInstruction/RedTeamingVLM},
- preview = {Li2024RedTV.png}
-}
-
-@article{Wang2023LargeLM,
- title = {Large Language Models are not Fair Evaluators},
- author = {Peiyi Wang and Lei Li and Liang Chen and Dawei Zhu and Binghuai Lin and Yunbo Cao and Qi Liu and Tianyu Liu and Zhifang Sui},
- journal = {In Proceedings of the
- Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2024},
- arxiv = {2305.17926},
- code = {https://github.com/i-Eval/FairEval},
- preview = {Wang2023LargeLM.png}
-}
-
-@article{Wang2023ACB,
- title = {A Challenging Benchmark for Low-Resource Learning},
- author = {Yudong Wang and Chang Ma and Qingxiu Dong and Lingpeng Kong and Jingjing Xu},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2024},
- arxiv = {2303.03840},
- code = {https://github.com/Qian2333/Hard-Bench},
- preview = {Wang2023ACB.png}
-}
-
-
-@article{Wang2024LoRAMD,
- title = {LoRA Meets Dropout under a Unified Framework},
- author = {Sheng Wang and Liheng Chen and Jiyue Jiang and Boyang Xue and Lingpeng Kong and Chuan Wu},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2024},
- arxiv = {2403.00812},
- preview = {Wang2024LoRAMD.png}
-}
-
-
-@article{Li2024GSMPlusAC,
- title = {GSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem Solvers},
- author = {Qintong Li and Leyang Cui and Xueliang Zhao and Lingpeng Kong and Wei Bi},
- journal = {In Proceedings of the
- Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2024},
- arxiv = {2402.19255},
- code = {https://qtli.github.io/GSM-Plus/},
- preview = {Li2024GSMPlusAC.png}
-}
-
-@article{Wang2024PRoLoRAPR,
- title = {PRoLoRA: Partial Rotation Empowers More Parameter-Efficient LoRA},
- author = {Sheng Wang and Boyang Xue and Jiacheng Ye and Jiyue Jiang and Liheng Chen and Lingpeng Kong and Chuan Wu},
- journal = {In Proceedings of the
- Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2024},
- arxiv = {2402.16902},
- preview = {Wang2024PRoLoRAPR.png}
-}
-
-@article{An2023LEvalIS,
- title = {L-Eval: Instituting Standardized Evaluation for Long Context Language Models},
- author = {Chenxin An and Shansan Gong and Ming Zhong and Mukai Li and Jun Zhang and Lingpeng Kong and Xipeng Qiu},
- journal = {In Proceedings of the
- Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2024},
- arxiv = {2307.11088},
- code = {https://github.com/OpenLMLab/LEval},
- preview = {An2023LEvalIS.png}
-}
-
-@article{Li2024MultimodalAA,
- title = {Multimodal ArXiv: A Dataset for Improving Scientific Comprehension of Large Vision-Language Models},
- author = {Lei Li and Yuqi Wang and Runxin Xu and Peiyi Wang and Xiachong Feng and Lingpeng Kong and Qi Liu},
- journal = {In Proceedings of the
- Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2024},
- arxiv = {2403.00231},
- code = {https://mm-arxiv.github.io/},
- preview = {Li2024MultimodalAA.png}
-}
-
-@article{Zheng2023SelfInfillingCG,
- title = {Self-Infilling Code Generation},
- author = {Lin Zheng and Jianbo Yuan and Zhi Zhang and Hongxia Yang and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2024},
- arxiv = {2311.17972},
- code = {https://github.com/LZhengisme/self-infilling},
- preview = {Zheng2023SelfInfillingCG.png}
-}
-
-@article{An2024TrainingFreeLS,
- title = {Training-Free Long-Context Scaling of Large Language Models},
- author = {Chenxin An and Fei Huang and Jun Zhang and Shansan Gong and Xipeng Qiu and Chang Zhou and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2024},
- arxiv = {2402.17463},
- code = {https://github.com/HKUNLP/ChunkLlama},
- preview = {An2024TrainingFreeLS.png}
-}
-
-@article{Zhao2023DecomposingTE,
- title = {Decomposing the Enigma: Subgoal-based Demonstration Learning for Formal Theorem Proving},
- author = {Xueliang Zhao and Wenda Li and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2024},
- arxiv = {2305.16366},
- code = {https://github.com/HKUNLP/subgoal-theorem-prover},
- preview = {Zhao2023DecomposingTE.png}
-}
-
-@article{Xu2023LemurHN,
- title = {Lemur: Harmonizing Natural Language and Code for Language Agents},
- author = {Yiheng Xu and Hongjin Su and Chen Xing and Boyu Mi and Qian Liu and Weijia Shi and Binyuan Hui and Fan Zhou and Yitao Liu and Tianbao Xie and Zhoujun Cheng and Siheng Zhao and Lingpeng Kong and Bailin Wang and Caiming Xiong and Tao Yu},
- journal = {In International Conference on Learning Representations (ICLR)},
- year = {2024},
- arxiv = {2310.06830},
- code = {https://github.com/OpenLemur/Lemur},
- preview = {Xu2023LemurHN.png}
-}
-
-@article{Yang2023UniTabEAU,
- title = {UniTabE: A Universal Pretraining Protocol for Tabular Foundation Model in Data Science},
- author = {Yazheng Yang and Yuqi Wang and Guangyi Liu and Ledell Yu Wu and Qi Liu},
- journal = {In International Conference on Learning Representations (ICLR)},
- year = {2024},
- arxiv = {2307.09249},
- preview = {Yang2023UniTabEAU.png}
-}
-
-@article{Li2023CanLM,
- title = {Can Language Models Understand Physical Concepts?},
- author = {Lei Li and Jingjing Xu and Qingxiu Dong and Ce Zheng and Qi Liu and Lingpeng Kong and Xu Sun},
- journal = {In Proceedings of the Conference on Empirical Methods in Natural Language
- Processing (EMNLP)},
- year = {2023},
- arxiv = {2305.14057},
- code = {https://github.com/TobiasLee/VEC},
- preview = {Li2023CanLM.png}
-}
-
-@article{Pi2023DetGPTDW,
- title = {DetGPT: Detect What You Need via Reasoning},
- author = {Renjie Pi and Jiahui Gao and Shizhe Diao and Rui Pan and Hanze Dong and Jipeng Zhang and Lewei Yao and Jianhua Han and Hang Xu and Lingpeng Kong Tong Zhang},
- journal = {In Proceedings of the Conference on Empirical Methods in Natural Language
- Processing (EMNLP)},
- year = {2023},
- arxiv = {2305.14167},
- code = {https://detgpt.github.io/},
- preview = {Pi2023DetGPTDW.png}
-}
-
-@article{Ye2023GeneratingDF,
- title = {Generating Data for Symbolic Language with Large Language Models},
- author = {Jiacheng Ye and Chengzu Li and Lingpeng Kong and Tao Yu},
- journal = {In Proceedings of the Conference on Empirical Methods in Natural Language
- Processing (EMNLP)},
- year = {2023},
- arxiv = {2305.13917},
- code = {https://github.com/HKUNLP/SymGen},
- preview = {Ye2023GeneratingDF.png}
-}
-
-@article{Gong2023DiffuSeqv2BD,
- title = {DiffuSeq-v2: Bridging Discrete and Continuous Text Spaces for Accelerated Seq2Seq Diffusion Models},
- author = {Shansan Gong and Mukai Li and Jiangtao Feng and Zhiyong Wu and Lingpeng Kong},
- journal = {In Findings of the Conference on Empirical Methods in
- Natural Language Processing (EMNLP Findings)},
- year = {2023},
- arxiv = {2310.05793},
- code = {https://github.com/Shark-NLP/DiffuSeq},
- preview = {Gong2023DiffuSeqv2BD.png}
-}
-
-@article{Zhao2023GIMLETAU,
- title = {GIMLET: A Unified Graph-Text Model for Instruction-Based Molecule Zero-Shot Learning},
- author = {Haiteng Zhao and Shengchao Liu and Chang Ma and Hannan Xu and Jie Fu and Zhihong Deng and Lingpeng Kong and Qi Liu},
- journal = {In Advances in Neural Information Processing Systems (NeurIPS)},
- year = {2023},
- arxiv = {2306.13089},
- code = {https://github.com/zhao-ht/GIMLET},
- preview = {Zhao2023GIMLETAU.png}
-}
-
-@article{Dong2023StatisticalKA,
- title = {Statistical Knowledge Assessment for Large Language Models},
- author = {Qingxiu Dong and Jingjing Xu and Lingpeng Kong and Zhifang Sui and Lei Li},
- journal = {In Advances in Neural Information Processing Systems (NeurIPS)},
- year = {2023},
- arxiv = {2305.10519},
- code = {https://github.com/dqxiu/KAssess},
- preview = {Dong2023StatisticalKA.png}
-}
-
-@article{Wang2022EvaluatingSL,
- title = {Evaluating Self-Supervised Learning for Molecular Graph Embeddings},
- author = {Hanchen Wang and Jean Kaddour and Shengchao Liu and Jian Tang and Matt J. Kusner and Joan Lasenby and Qi Liu},
- journal = {In Advances in Neural Information Processing Systems (NeurIPS)},
- year = {2023},
- volume = {abs/2206.08005},
- url = {https://api.semanticscholar.org/CorpusID:249712002},
- preview = {Wang2022EvaluatingSL.png}
-}
-
-@article{Wu2022SelfAdaptiveIL,
- title = {Self-Adaptive In-Context Learning: An Information Compression Perspective for In-Context Example Selection and Ordering},
- author = {Zhiyong Wu and Yaoxiang Wang and Jiacheng Ye and Lingpeng Kong},
- journal = {In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2023},
- arxiv = {2212.10375},
- code = {https://github.com/Shark-NLP/self-adaptive-ICL},
- preview = {Wu2022SelfAdaptiveIL.png}
-}
-
-@article{Li2022ExplanationRV,
- title = {Explanation Regeneration via Information Bottleneck},
- author = {Qintong Li and Zhiyong Wu and Lingpeng Kong and Wei Bi},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2023},
- arxiv = {2212.09603},
- code = {https://github.com/qtli/EIB},
- preview = {Li2022ExplanationRV.png}
-}
-
-@article{Jiang2023ACS,
- title = {A Cognitive Stimulation Dialogue System with Multi-source Knowledge Fusion for Elders with Cognitive Impairment},
- author = {Jiyue Jiang and Sheng Wang and Qintong Li and Lingpeng Kong and Chuan Wu},
- journal = {In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2023},
- arxiv = {2305.08200},
- preview = {Jiang2023ACS.png}
-}
-
-@article{Zhao2023SORTIEDS,
- title = {SORTIE: Dependency-Aware Symbolic Reasoning for Logical Data-to-text Generation},
- author = {Xueliang Zhao and Tingchen Fu and Lemao Liu and Lingpeng Kong and Shuming Shi and Rui Yan},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2023},
- pdf = {https://aclanthology.org/2023.findings-acl.715.pdf},
- preview = {Zhao2023SORTIEDS.png}
-}
-
-@article{Su2022OneEA,
- title = {One Embedder, Any Task: Instruction-Finetuned Text Embeddings},
- author = {Hongjin Su and Weijia Shi and Jungo Kasai and Yizhong Wang and Yushi Hu and Mari Ostendorf and Wen-tau Yih and Noah A. Smith and Luke Zettlemoyer and Tao Yu},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2023},
- arxiv = {2212.09741},
- code = {https://instructor-embedding.github.io/},
- preview = {Su2022OneEA.png}
-}
-
-@article{Ye2023CompositionalEF,
- title = {Compositional Exemplars for In-context Learning},
- author = {Jiacheng Ye and Zhiyong Wu and Jiangtao Feng and Tao Yu and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2023},
- arxiv = {2302.05698},
- code = {https://github.com/HKUNLP/icl-ceil},
- preview = {Ye2023CompositionalEF.png}
-}
-
-@article{Zhang2022CABCA,
- title = {CAB: Comprehensive Attention Benchmarking on Long Sequence Modeling},
- author = {Jinchao Zhang and Shuyang Jiang and Jiangtao Feng and Lin Zheng and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2023},
- arxiv = {2210.07661},
- code = {https://github.com/Shark-NLP/CAB},
- preview = {Zhang2022CABCA.png}
-}
-
-@article{Lai2022DS1000AN,
- title = {DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation},
- author = {Yuhang Lai and Chengxi Li and Yiming Wang and Tianyi Zhang and Ruiqi Zhong and Luke Zettlemoyer and Scott Yih and Daniel Fried and Si-yi Wang and Tao Yu},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2023},
- arxiv = {2211.11501},
- code = {https://github.com/xlang-ai/DS-1000},
- preview = {Lai2022DS1000AN.png}
-}
-
-@article{Zheng2023EfficientAV,
- title = {Efficient Attention via Control Variates},
- author = {Lin Zheng and Jianbo Yuan and Chong Wang and Lingpeng Kong},
- journal = {In International Conference on Learning Representations (ICLR)},
- year = {2023},
- arxiv = {2302.04542},
- code = {https://github.com/HKUNLP/efficient-attention},
- preview = {Zheng2023EfficientAV.png}
-}
-
-@article{gao2023selfguided,
- title = {Self-Guided Noise-Free Data Generation for Efficient Zero-Shot Learning},
- author = {Jiahui Gao and Renjie Pi and Lin Yong and Hang Xu and Jiacheng Ye and Zhiyong Wu and Weizhong Zhang and Xiaodan Liang and Zhenguo Li and Lingpeng Kong},
- journal = {International Conference on Learning Representations (ICLR)},
- location = {Kigali, Rwanda},
- year = {2023},
- arxiv = {2205.12679},
- code = {https://github.com/SumilerGAO/SunGen},
- preview = {gao2023selfguided.png}
-}
-
-@article{Qin2023ToeplitzNN,
- title = {Toeplitz Neural Network for Sequence Modeling},
- author = {Zhen Qin and Xiaodong Han and Weixuan Sun and Bowen He and Dong Li and Dongxu Li and Yuchao Dai and Lingpeng Kong and Yiran Zhong},
- journal = {International Conference on Learning Representations (ICLR)},
- year = {2023},
- arxiv = {2305.04749},
- code = {https://github.com/OpenNLPLab/Tnn},
- preview = {Qin2023ToeplitzNN.png}
-}
-
-@article{Gong2022DiffuSeqST,
- title = {DiffuSeq: Sequence to Sequence Text Generation with Diffusion Models},
- author = {Shansan Gong and Mukai Li and Jiangtao Feng and Zhiyong Wu and Lingpeng Kong},
- journal = {International Conference on Learning Representations (ICLR)},
- year = {2023},
- arxiv = {2210.08933},
- code = {https://github.com/Shark-NLP/DiffuSeq},
- preview = {Gong2022DiffuSeqST.png}
-}
-
-@article{cheng2023binding,
- title = {Binding Language Models in Symbolic Languages},
- author = {Cheng, Zhoujun and Xie, Tianbao and Shi, Peng and Li, Chengzu and Nadkarni, Rahul and Hu, Yushi and Xiong, Caiming and Radev, Dragomir and Ostendorf, Mari and Zettlemoyer, Luke and Smith, Noah A and Yu, Tao},
- journal = {International Conference on Learning Representations (ICLR)},
- location = {Berlin, Germany},
- year = {2023},
- arxiv = {2210.02875},
- code = {https://lm-code-binder.github.io/},
- poster = {https://lm-code-binder.github.io/},
- selected = {y},
- preview = {Gong2022DiffuSeqST.png}
-}
-
-@article{su2023selective,
- title = {Selective Annotation Makes Language Models Better Few-Shot Learners},
- author = {Su, Hongjin and Kasai, Jungo and Wu, Chen Henry and Shi, Weijia and Wang, Tianlu and Xin, Jiayi and Zhang, Rui and Ostendorf, Mari and Zettlemoyer, Luke and Smith, Noah A. and Yu, Tao},
- journal = {International Conference on Learning Representations (ICLR)},
- year = {2023},
- arxiv = {2209.01975},
- code = {https://github.com/HKUNLP/icl-selective-annotation},
- preview = {su2023selective.png}
-}
-
-@article{chen2023unsupervised,
- title = {Unsupervised Explanation Generation via Correct Instantiations},
- author = {Sijie Chen and Zhiyong Wu and Jiangjie Chen and Zhixing Li and Yang Liu and Lingpeng Kong},
- journal = {In Proceedings of AAAI Conference on Artificial Intelligence (AAAI)},
- location = {Washington, DC},
- year = {2023},
- arxiv = {2211.11160},
- code = {https://github.com/Shark-NLP/Neon},
- preview = {chen2023unsupervised.png}
-}
-
-@article{Wang2022AnES,
- title = {An Empirical Study of Retrieval-Enhanced Graph Neural Networks},
- author = {Dingmin Wang and Shengchao Liu and Hanchen Wang and Bernardo Cuenca Grau and Linfeng Song and Jian Tang and Song Le and Qi Li},
- journal = {European Conference on Artificial Intelligence (ECAI)},
- year = {2023},
- preview = {Wang2022AnES.png}
-}
-
-@article{ma2023retrieved,
- title = {Retrieved Sequence Augmentation for Protein Representation Learning},
- author = {Chang Ma and Haiteng Zhao and Lin Zheng and Jiayi Xin and Qintong Li and Lijun Wu and Zhihong Deng and Yang Lu and Qi Liu and Lingpeng Kong},
- journal = {arXiv preprint},
- year = {2023},
- arxiv = {2302.12563},
- code = {https://github.com/HKUNLP/RSA},
- preview = {ma2023retrieved.png}
-}
-
-
-
-
-%%%%%%%%%%%% 2022 %%%%%%%%%%%%
-
-
-@article{Ye2022ProGenPZ,
- title = {ProGen: Progressive Zero-shot Dataset Generation via In-context Feedback},
- author = {Jiacheng Ye and Jiahui Gao and Jiangtao Feng and Zhiyong Wu and Tao Yu and Lingpeng Kong},
- journal = {In Findings of the Conference on Empirical Methods in Natural Language Processing (EMNLP Findings)},
- year = {2022},
- arxiv = {2210.12329},
- code = {https://github.com/HKUNLP/ProGen},
- preview = {Ye2022ProGenPZ.png}
-}
-
-@article{liu2022augmenting,
- title = {Augmenting Multi-Turn Text-to-SQL Datasets with Self-Play},
- author = {Liu, Qi and Ye, Zihuiwen and Yu, Tao and Blunsom, Phil and Song, Linfeng},
- journal = {In Findings of the Conference on Empirical Methods in Natural Language Processing (EMNLP Findings)},
- year = {2022},
- arxiv = {2210.12096},
- code = {https://github.com/leuchine/self_play_picard},
- preview = {liu2022augmenting.png}
-}
-
-@article{ye2022zerogen,
- title = {ZeroGen: Efficient Zero-shot Learning via Dataset Generation},
- author = {Ye, Jiacheng and Gao, Jiahui and Li, Qintong and Xu, Hang and Feng, Jiangtao and Wu, Zhiyong and Yu, Tao and Kong, Lingpeng},
- journal = {In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)},
- year = {2022},
- arxiv = {2202.07922},
- code = {https://github.com/jiacheng-ye/ZeroGen},
- preview = {ye2022zerogen.png}
-}
-
-@article{xie2022unifiedskg,
- title = {UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models},
- author = {Xie, Tianbao and Wu, Chen Henry and Shi, Peng and Zhong, Ruiqi and Scholak, Torsten and Yasunaga, Michihiro and Wu, Chien-Sheng and Zhong, Ming and Yin, Pengcheng and Wang, Sida and Zhong, Victor and Wang, Bailin and Li, Chengzu and Boyle, Connor and Ni, Ansong and Yao, Ziyu and Radev, Dragomir and Xiong, Caiming and Kong, Lingpeng and Zhang, Rui and Smith, Noah A. and Zettlemoyer, Luke and Yu, Tao},
- journal = {In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)},
- year = {2022},
- arxiv = {2201.05966},
- code = {https://github.com/hkunlp/unifiedskg},
- poster = {https://unifiedskg.com/},
- preview = {xie2022unifiedskg.png}
-}
-
-@article{An2022CoNTCN,
- title = {CoNT: Contrastive Neural Text Generation},
- author = {Chenxin An and Jiangtao Feng and Kai Lv and Lingpeng Kong and Xipeng Qiu and Xuanjing Huang},
- journal = {In Advances in Neural Information Processing Systems (NeurIPS)},
- year = {2022},
- arxiv = {2205.14690},
- code = {https://github.com/Shark-NLP/CoNT},
- preview = {An2022CoNTCN.png}
-}
-
-@article{zheng2022linear,
- title = {Linear Complexity Randomized Self-attention Mechanism},
- author = {Lin Zheng and Chong Wang and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2022},
- arxiv = {2204.04667},
- code = {https://github.com/HKUNLP/efficient-attention},
- preview = {zheng2022linear.png}
-}
-
-@article{zheng2022ripple,
- title = {Ripple Attention for Visual Perception with Sub-quadratic Complexity},
- author = {Lin Zheng and Huijie Pan and Lingpeng Kong},
- journal = {In Proceedings of the International Conference on Machine Learning (ICML)},
- year = {2022},
- arxiv = {2110.02453},
- preview = {zheng2022ripple.png}
-}
-
-@article{li2022event,
- title = {Event Transition Planning for Open-ended Text Generation},
- author = {Qintong Li and Piji Li and Wei Bi and Zhaochun Ren and Yuxuan Lai and Lingpeng Kong},
- journal = {In Findings of the Annual Meeting of the Association for Computational Linguistics (ACL Findings)},
- year = {2022},
- arxiv = {2204.09453},
- code = {https://github.com/qtli/EventPlanforTextGen},
- preview = {li2022event.png}
-}
-
-
-@article{wu2022lexical,
- title = {Lexical Knowledge Internalization for Neural Dialog Generation},
- author = {Zhiyong Wu and Wei Bi and Xiang Li and Lingpeng Kong and Ben Kao},
- journal = {In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2022},
- arxiv = {2205.01941},
- code = {https://github.com/LividWo/KI},
- preview = {wu2022lexical.png}
-}
-
-@inproceedings{Prange2021LinguisticFG,
- title = {Linguistic Frameworks Go Toe-to-Toe at Neuro-Symbolic Language Modeling},
- author = {Jakob Prange and Nathan Schneider and Lingpeng Kong},
- booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)},
- year = {2022},
- arxiv = {2112.07874},
- preview = {wu2022lexical.png}
-}
-
-
-%%%%%%%%%%%% 2021 %%%%%%%%%%%%%%%%%%%
-
-@article{zheng2021cascaded,
- title = {Cascaded Head-colliding Attention},
- author = {Lin Zheng and Zhiyong Wu and Lingpeng Kong},
- journal = {In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2021},
- arxiv = {2105.14850},
- code = {https://github.com/LZhengisme/CODA},
- preview = {zheng2021cascaded.png}
-}
-
-@article{wu2021good,
- title = {Good for Misconceived Reasons: An Empirical Revisiting on the Need for Visual Context in Multimodal Machine Translation},
- author = {Zhiyong Wu and Lingpeng Kong and Wei Bi and Xiang Li and Ben Kao},
- journal = {In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)},
- year = {2021},
- arxiv = {2105.14462},
- code = {https://github.com/LividWo/Revisit-MMT},
- preview = {wu2021good.png}
-}
diff --git a/_config.yml b/_config.yml
deleted file mode 100644
index 0a7d8d1..0000000
--- a/_config.yml
+++ /dev/null
@@ -1,329 +0,0 @@
-# -----------------------------------------------------------------------------
-# Site settings
-# -----------------------------------------------------------------------------
-
-title: blank # the website title (if blank, full name will be used instead)
-first_name: HKU NLP Group
-middle_name:
-last_name:
-email:
-description: > # the ">" symbol means to ignore newlines until "footer_text:"
- A simple, whitespace theme for academics. Based on [*folio](https://github.com/bogoli/-folio) design.
-footer_text: >
- Powered by Jekyll with al-folio theme.
- Hosted by GitHub Pages.
-keywords: # jekyll, jekyll-theme, academic-website, portfolio-website # add your own keywords or leave empty
-
-
-lang: en # the language of your site (for example: en, fr, cn, ru, etc.)
-icon: icon.png # the emoji used as the favicon (alternatively, provide image name in /assets/img/)
-
-url: https://hkunlp.github.io # the base hostname & protocol for your site
-baseurl: # the subpath of your site, e.g. /blog/
-last_updated: false # set to true if you want to display last updated in the footer
-impressum_path: # set to path to include impressum link in the footer, use the same path as permalink in a page, helps to conform with EU GDPR
-
-# -----------------------------------------------------------------------------
-# RSS Feed
-# -----------------------------------------------------------------------------
-# will use title and url fields
-# Take a look to https://github.com/jekyll/jekyll-feed for more customization
-
-rss_icon: false
-
-# -----------------------------------------------------------------------------
-# Layout
-# -----------------------------------------------------------------------------
-
-navbar_fixed: true
-footer_fixed: true
-
-# Dimensions
-max_width: 800px
-
-# TODO: add layout settings (single page vs. multi-page)
-
-# -----------------------------------------------------------------------------
-# Open Graph & Schema.org
-# -----------------------------------------------------------------------------
-# Display links to the page with a preview object on social media.
-serve_og_meta: false # Include Open Graph meta tags in the HTML head
-serve_schema_org: false # Include Schema.org in the HTML head
-og_image: # The site-wide (default for all links) Open Graph preview image
-
-# -----------------------------------------------------------------------------
-# Social integration
-# -----------------------------------------------------------------------------
-
-github_username: HKUNLP # your GitHub user name
-gitlab_username: # your GitLab user name
-twitter_username: # your Twitter handle
-linkedin_username: # your LinkedIn user name
-scholar_userid: # your Google Scholar ID
-orcid_id: # your ORCID ID
-medium_username: # your Medium username
-quora_username: # your Quora username
-publons_id: # your ID on Publons
-research_gate_profile: # your profile on ResearchGate
-blogger_url: # your blogger URL
-work_url: # work page URL
-keybase_username: # your keybase user name
-wikidata_id: # your wikidata id
-dblp_url: # your DBLP profile url
-stackoverflow_id: #your stackoverflow id
-
-contact_note: >
- You can even add a little note about which of these is the best way to reach you.
-
-# -----------------------------------------------------------------------------
-# Analytics and search engine verification
-# -----------------------------------------------------------------------------
-
-google_analytics: # your google-analytics ID (format: UA-XXXXXXXXX)
-panelbear_analytics: # panelbear analytics site ID (format: XXXXXXXXX)
-
-google_site_verification: # your google-site-verification ID (Google Search Console)
-bing_site_verification: # out your bing-site-verification ID (Bing Webmaster)
-
-# -----------------------------------------------------------------------------
-# Blog
-# -----------------------------------------------------------------------------
-
-blog_name: Blog # blog_name will be displayed in your blog page
-blog_nav_title: Blog # your blog must have a title for it to be displayed in the nav bar
-blog_description: Posts on research projects @ HKU NLP Group
-permalink: /blog/:year/:title/
-
-# Pagination
-pagination:
- enabled: true
-
-# Comments
-disqus_shortname: al-folio # put your disqus shortname
-# https://help.disqus.com/en/articles/1717111-what-s-a-shortname
-
-# External sources.
-# If you have blog posts published on medium.com or other exteranl sources,
-# you can display them in your blog by adding a link to the RSS feed.
-# external_sources:
-# - name: medium.com
-# rss_url: https://medium.com/@al-folio/feed
-
-# -----------------------------------------------------------------------------
-# Collections
-# -----------------------------------------------------------------------------
-
-collections:
- news:
- defaults:
- layout: post
- output: true
- permalink: /news/:path/
- projects:
- output: true
- permalink: /projects/:path/
- people:
- output: true
- permalink: /people/:path/
-
-news_scrollable: true # adds a vertical scroll bar if there are more than 3 news items
-news_limit: 5 # leave blank to include all the news in the `_news` folder
-
-# -----------------------------------------------------------------------------
-# Jekyll settings
-# -----------------------------------------------------------------------------
-
-# Markdown and syntax highlight
-markdown: kramdown
-highlighter: rouge
-highlight_theme_light: github # https://github.com/jwarby/jekyll-pygments-themes
-highlight_theme_dark: native # https://github.com/jwarby/jekyll-pygments-themes
-kramdown:
- input: GFM
- syntax_highlighter_opts:
- css_class: 'highlight'
- span:
- line_numbers: false
- block:
- line_numbers: false
- start_line: 1
-
-# Includes & excludes
-include: ['_pages']
-exclude:
- - bin
- - Gemfile
- - Gemfile.lock
- - vendor
-keep_files:
- - CNAME
- - .nojekyll
- - .git
-
-# Plug-ins
-plugins:
- - jekyll-archives
- - jekyll-diagrams
- - jekyll-email-protect
- - jekyll-feed
- - jekyll-imagemagick
- - jekyll-minifier
- - jekyll-paginate-v2
- - jekyll/scholar
- - jekyll-sitemap
- - jekyll-target-blank
- - jekyll-twitter-plugin
- - jekyll-redirect-from
- - jemoji
- - jekyll-loading-lazy
-
-# Sitemap settings
-defaults:
- - scope:
- path: "assets/**/*.*"
- values:
- sitemap: false
-
-# -----------------------------------------------------------------------------
-# Jekyll Minifier
-# -----------------------------------------------------------------------------
-
-jekyll-minifier:
- exclude: ['robots.txt']
- uglifier_args:
- harmony: true
-
-# -----------------------------------------------------------------------------
-# Jekyll Archives
-# -----------------------------------------------------------------------------
-
-jekyll-archives:
- enabled: [year, tags, categories] # enables year, tag and category archives (remove if you need to disable one of them).
- layouts:
- year: archive-year
- tag: archive-tag
- category: archive-category
- permalinks:
- year: '/blog/:year/'
- tag: '/blog/tag/:name/'
- category: '/blog/category/:name/'
-
-display_tags: ['attention', 'transformers', 'generation', 'multi-modality', 'NLP'] # this tags will be dispalyed on the front page of your blog
-
-# -----------------------------------------------------------------------------
-# Jekyll Scholar
-# -----------------------------------------------------------------------------
-
-scholar:
-
- last_name: Einstein
- first_name: [Albert, A.]
-
- style: apa
- locale: en
-
- source: /_bibliography/
- bibliography: papers.bib
- bibliography_template: bib
- # Note: if you have latex math in your bibtex, the latex filter
- # preprocessing may conflict with MathJAX if the latter is enabled.
- # See https://github.com/alshedivat/al-folio/issues/357.
- bibtex_filters: [latex, smallcaps, superscript]
-
- replace_strings: true
- join_strings: true
-
- details_dir: bibliography
- details_layout: bibtex.html
- details_link: Details
-
- query: "@*"
-
-# Filter out certain bibtex entry keywords used internally from the bib output
-filtered_bibtex_keywords: [abbr, abstract, arxiv, bibtex_show, html, pdf, selected, supp, blog, code, poster, slides, website, preview]
-
-# Maximum number of authors to be shown, other authors will be visible on hover, leave blank to show all authors
-max_author_limit: 12
-
-# -----------------------------------------------------------------------------
-# Responsive WebP Images
-# -----------------------------------------------------------------------------
-
-imagemagick:
- enabled: true # enables responsive images for your site (recomended, see https://github.com/alshedivat/al-folio/issues/537)
- widths:
- - 480
- - 800
- - 1400
- input_directories:
- - assets/img/
- input_formats:
- - ".jpg"
- - ".jpeg"
- - ".png"
- - ".tiff"
- output_formats:
- webp: "-quality 75%"
-
-# -----------------------------------------------------------------------------
-# Jekyll Diagrams
-# -----------------------------------------------------------------------------
-
-jekyll-diagrams:
- # configuration, see https://github.com/zhustec/jekyll-diagrams.
- # feel free to comment out this section if not using jekyll diagrams.
-
-
-# -----------------------------------------------------------------------------
-# Optional Features
-# -----------------------------------------------------------------------------
-
-enable_google_analytics: false # enables google analytics
-enable_panelbear_analytics: false # enables panelbear analytics
-enable_google_verification: false # enables google site verification
-enable_bing_verification: false # enables bing site verification
-enable_masonry: true # enables automatic project cards arangement
-enable_math: true # enables math typesetting (uses MathJax)
-enable_tooltips: false # enables automatic tooltip links generated
- # for each section titles on pages and posts
-enable_darkmode: true # enables switching between light/dark modes
-enable_navbar_social: false # enables displaying social links in the
- # navbar on the about page
-enable_project_categories: true # enables categorization of projects into
- # multiple categories
-enable_people_categories: true # enables categorization of researchers into
- # multiple categories
-enable_medium_zoom: false # enables image zoom feature (as on medium.com)
-
-
-# -----------------------------------------------------------------------------
-# Library versions
-# -----------------------------------------------------------------------------
-
-academicons:
- version: "1.9.1"
- integrity: "sha256-i1+4qU2G2860dGGIOJscdC30s9beBXjFfzjWLjBRsBg="
-bootstrap:
- version: "4.6.1"
- integrity:
- css: "sha256-DF7Zhf293AJxJNTmh5zhoYYIMs2oXitRfBjY+9L//AY="
- js: "sha256-fgLAgv7fyCGopR/gBNq2iW3ZKIdqIcyshnUULC4vex8="
-fontawesome:
- version: "5.15.4"
- integrity: "sha256-mUZM63G8m73Mcidfrv5E+Y61y7a12O5mW4ezU3bxqW4="
-jquery:
- version: "3.6.0"
- integrity: "sha256-/xUj+3OJU5yExlq6GSYGSHk7tPXikynS7ogEvDej/m4="
-mathjax:
- version: "3.2.0"
-masonry:
- version: "4.2.2"
- integrity: "sha256-Nn1q/fx0H7SNLZMQ5Hw5JLaTRZp0yILA/FRexe19VdI="
-mdb:
- version: "4.20.0"
- integrity:
- css: "sha256-jpjYvU3G3N6nrrBwXJoVEYI/0zw8htfFnhT9ljN3JJw="
- js: "sha256-NdbiivsvWt7VYCt6hYNT3h/th9vSTL4EDWeGs5SN3DA="
-medium_zoom:
- version: "1.0.6"
- integrity: "sha256-EdPgYcPk/IIrw7FYeuJQexva49pVRZNmt3LculEr7zM="
diff --git a/_data/coauthors.yml b/_data/coauthors.yml
deleted file mode 100644
index 8ed5212..0000000
--- a/_data/coauthors.yml
+++ /dev/null
@@ -1,34 +0,0 @@
-"Adams":
- - firstname: ["Edwin", "E.", "E. P.", "Edwin Plimpton"]
- url: https://en.wikipedia.org/wiki/Edwin_Plimpton_Adams
-
-"Podolsky":
- - firstname: ["Boris", "B.", "B. Y.", "Boris Yakovlevich"]
- url: https://en.wikipedia.org/wiki/Boris_Podolsky
-
-"Rosen":
- - firstname: ["Nathan", "N."]
- url: https://en.wikipedia.org/wiki/Nathan_Rosen
-
-"Bach":
- - firstname: ["Johann Sebastian", "J. S."]
- url: https://en.wikipedia.org/wiki/Johann_Sebastian_Bach
-
- - firstname: ["Carl Philipp Emanuel", "C. P. E."]
- url: https://en.wikipedia.org/wiki/Carl_Philipp_Emanuel_Bach
-
-"Przibram":
- - firstname: ["Karl"]
- url: https://link.springer.com/article/10.1007/s00016-019-00242-z
-
-"Schrödinger":
- - firstname: ["Erwin"]
- url: https://en.wikipedia.org/wiki/Erwin_Schr%C3%B6dinger
-
-"Lorentz":
- - firstname: ["Hendrik Antoon"]
- url: https://en.wikipedia.org/wiki/Hendrik_Lorentz
-
-"Planck":
- - firstname: ["Max"]
- url: https://en.wikipedia.org/wiki/Max_Planck
diff --git a/_data/cv.yml b/_data/cv.yml
deleted file mode 100644
index 5b11572..0000000
--- a/_data/cv.yml
+++ /dev/null
@@ -1,97 +0,0 @@
-- title: General Information
- type: map
- contents:
- - name: Full Name
- value: Albert Einstein
- - name: Date of Birth
- value: 14th March 1879
- - name: Languages
- value: English, German
-
-- title: Education
- type: time_table
- contents:
- - title: PhD
- institution: University of Zurich, Zurich, Switzerland
- year: 1905
- description:
- - Description 1.
- - Description 2.
- - title: Description 3.
- contents:
- - Sub-description 1.
- - Sub-description 2.
- - title: Federal teaching diploma
- institution: Eidgenössische Technische Hochschule, Zurich, Switzerland
- year: 1900
- description:
- - Description 1.
- - Description 2.
-
-- title: Experience
- type: time_table
- contents:
- - title: Professor of Theoretical Physics
- institution: Institute for Advanced Study, Princeton University
- year: 1933 - 1955
- description:
- - Description 1.
- - Description 2.
- - title: Description 3.
- contents:
- - Sub-description 1.
- - Sub-description 2.
- - title: Visiting Professor
- institution: California Institute of Technology, Pasadena, California, US
- year: 1933
- description:
- - Description 1.
- - Description 2.
-
- - title: Director
- institution: Kaiser Wilhelm Institute for Physics, Berlin, Germany.
- year: 1917-1933
-
- - title: Professor of Theoretical Physics
- institution: Karl-Ferdinand University, Prague, Czechoslovakia
- year: 1911 - 1917
- description:
-
- - title: Associate Professor of Theoretical Physics
- institution: University of Zurich, Zurich, Switzerland
- year: 1909 - 1911
-
-- title: Open Source Projects
- type: time_table
- contents:
- - title: al-folio
- year: 2015-now
- description: A beautiful, simple, clean, and responsive Jekyll theme for academics.
-
-- title: Honors and Awards
- type: time_table
- contents:
- - year: 1921
- items:
- - Nobel Prize in Physics
- - Matteucci Medal
- - year: 2029
- items:
- - Max Planck Medal
-
-- title: Academic Interests
- type: nested_list
- contents:
- - title: Topic 1.
- items:
- - Description 1.
- - Description 2.
- - title: Topic 2.
- items:
- - Description 1.
- - Description 2.
-
-- title: Other Interests
- type: list
- contents:
- - Hobbies: Hobby 1, Hobby 2, etc.
diff --git a/_includes/cv/list.html b/_includes/cv/list.html
deleted file mode 100644
index 7562585..0000000
--- a/_includes/cv/list.html
+++ /dev/null
@@ -1,5 +0,0 @@
-
- {% for content in entry.contents %}
-
{{ content }}
- {% endfor %}
-
\ No newline at end of file
diff --git a/_includes/cv/map.html b/_includes/cv/map.html
deleted file mode 100644
index e0d1983..0000000
--- a/_includes/cv/map.html
+++ /dev/null
@@ -1,8 +0,0 @@
-
- {% for content in entry.contents %}
-
-
{{ content.name }}
-
{{ content.value }}
-
- {% endfor %}
-
\ No newline at end of file
diff --git a/_includes/cv/nested_list.html b/_includes/cv/nested_list.html
deleted file mode 100644
index 4778aca..0000000
--- a/_includes/cv/nested_list.html
+++ /dev/null
@@ -1,14 +0,0 @@
-
- {% for content in entry.contents %}
-
-
{{ content.title }}
- {% if content.items %}
-
- {% for subitem in content.items %}
-
{{ subitem }}
- {% endfor %}
-
- {% endif %}
-
- {% endfor %}
-
\ No newline at end of file
diff --git a/_includes/cv/time_table.html b/_includes/cv/time_table.html
deleted file mode 100644
index 123b9d0..0000000
--- a/_includes/cv/time_table.html
+++ /dev/null
@@ -1,59 +0,0 @@
-
- {% for content in entry.contents %}
-
-
- {% if content.year %}
-
-
- {{ content.year }}
-
-
- {% endif %}
-
- {% if content.title %}
-
{{content.title}}
- {% endif %}
- {% if content.institution %}
-
{{content.institution}}
- {% endif %}
- {% if content.description %}
-
- {% for item in content.description %}
-
- {% if item.contents %}
- {{ item.title }}
-
- {% for subitem in item.contents %}
-
{{ subitem }}
- {% endfor %}
-
- {% else %}
- {{ item }}
- {% endif %}
-
- {% endfor %}
-
- {% endif %}
- {% if content.items %}
-
- {% for item in content.items %}
-
- {% if item.contents %}
- {{ item.title }}
-
- {% for subitem in item.contents %}
-
{{ subitem }}
- {% endfor %}
-
- {% else %}
- {{ item }}
- {% endif %}
-
- {% endfor %}
-
- {% endif %}
-
-
-
- {% endfor %}
-
\ No newline at end of file
diff --git a/_includes/figure.html b/_includes/figure.html
deleted file mode 100644
index ffda4d1..0000000
--- a/_includes/figure.html
+++ /dev/null
@@ -1,26 +0,0 @@
-{%- assign img_path = include.path | remove: ".jpg" | remove: ".jpeg" | remove: ".png" | remove: ".tiff" -%}
-
-
-
-
- {% if site.imagemagick.enabled %}
- {% for i in site.imagemagick.widths -%}
-
- {% endfor -%}
- {% endif %}
-
-
-
-
-
- {%- if include.caption -%}{{ include.caption }}{%- endif %}
-
-
\ No newline at end of file
diff --git a/_includes/footer.html b/_includes/footer.html
deleted file mode 100644
index 721d262..0000000
--- a/_includes/footer.html
+++ /dev/null
@@ -1,27 +0,0 @@
-{% if site.footer_fixed %}
-
-{%- else -%}
-
-{%- endif %}
\ No newline at end of file
diff --git a/_includes/head.html b/_includes/head.html
deleted file mode 100644
index 1661581..0000000
--- a/_includes/head.html
+++ /dev/null
@@ -1,31 +0,0 @@
-
- {% include metadata.html %}
-
-
-
-
-
-
-
-
-
-
-
-
-
-
- {% if site.icon.size < 3 %}
-
- {% elsif site.icon != blank %}
-
- {% endif %}
-
-
-
-
- {% if site.enable_darkmode %}
-
-
-
-
- {% endif %}
diff --git a/_includes/header.html b/_includes/header.html
deleted file mode 100644
index 092a046..0000000
--- a/_includes/header.html
+++ /dev/null
@@ -1,112 +0,0 @@
-
-
-
-
-
-
\ No newline at end of file
diff --git a/_includes/metadata.html b/_includes/metadata.html
deleted file mode 100644
index ce58cc3..0000000
--- a/_includes/metadata.html
+++ /dev/null
@@ -1,156 +0,0 @@
-{% if site.enable_google_verification or site.enable_bing_verification %}
-
- {% if site.enable_google_verification -%}
-
- {%- endif -%}
- {% if site.enable_bing_verification -%}
-
- {%- endif -%}
-{%- endif %}
-
-
-
-
-
-
- {%- if site.title == "blank" -%}
- {%- capture title -%}{{ site.first_name }} {{ site.middle_name }} {{ site.last_name }}{%- endcapture -%}
- {%- else -%}
- {%- capture title -%}{{ site.title }}{%- endcapture -%}
- {%- endif -%}
- {% if page.url == '/blog/index.html' %}
- {{ site.blog_nav_title }} | {{ title }}
- {%- elsif page.title != "blank" and page.url != "/" -%}
- {{ page.title }} | {{ title }}
- {%- else -%}
- {{ title }}
- {%- endif -%}
-
-
-
-{%- if page.keywords or site.keywords %}
-
-{%- endif %}
-
-{%- if site.serve_og_meta %}
-
-
-
-
-
-
-
- {% if page.og_image or site.og_image -%}
-
- {%- endif %}
-
-
-
-
-
-
- {% if page.og_image or site.og_image -%}
-
- {%- endif %}
- {% if site.twitter_username -%}
-
-
- {%- endif %}
-{%- endif %}
-
-{%- if site.serve_schema_org %}
-
-
- {%- comment -%} Social links generator for "sameAs schema" {%- endcomment %}
- {% assign sameaslinks = "" | split: "," %}
- {%- if site.orcid_id -%}
- {%- capture link -%}https://orcid.org/{{ site.orcid_id }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.scholar_userid -%}
- {%- capture link -%}https://scholar.google.com/citations?user={{ site.scholar_userid }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.publons_id -%}
- {%- capture link -%}https://publons.com/a/{{ site.publons_id }}/{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.research_gate_profile -%}
- {%- capture link -%}https://www.researchgate.net/profile/{{site.research_gate_profile}}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.github_username -%}
- {%- capture link -%}https://github.com/{{ site.github_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.linkedin_username -%}
- {%- capture link -%}https://www.linkedin.com/in/{{ site.linkedin_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.twitter_username -%}
- {%- capture link -%}https://twitter.com/{{ site.twitter_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.medium_username -%}
- {%- capture link -%}https://medium.com/@{{ site.medium_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.quora_username -%}
- {%- capture link -%}https://www.quora.com/profile/{{ site.quora_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.blogger_url -%}
- {%- capture link -%}{{ site.blogger_url }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.work_url -%}
- {%- capture link -%}{{ site.work_url }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.wikidata_id -%}
- {%- capture link -%}https://www.wikidata.org/wiki/{{ site.wikidata_id }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.strava_userid -%}
- {%- capture link -%}https://www.strava.com/athletes/{{ site.strava_userid }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.keybase_username -%}
- {%- capture link -%}https://keybase.io/{{ site.keybase_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.gitlab_username -%}
- {%- capture link -%}https://gitlab.com/{{ site.gitlab_username }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.dblp_url -%}
- {%- capture link -%}{{ site.dblp_url }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if site.stackoverflow_id -%}
- {%- capture link -%}https://stackoverflow.com/users/{{ site.stackoverflow_id }}{%- endcapture -%}
- {%- assign sameaslinks = sameaslinks | push: link -%}
- {%- endif -%}
- {%- if sameaslinks != blank -%}
- {%- assign sameaslinks = sameaslinks | split: "" -%}
- {%- endif -%}
-
-
-{%- endif %}
diff --git a/_includes/news.html b/_includes/news.html
deleted file mode 100644
index 307e532..0000000
--- a/_includes/news.html
+++ /dev/null
@@ -1,31 +0,0 @@
-
-
- {%- for project in sorted_projects -%}
- {% include projects_horizontal.html %}
- {%- endfor %}
-
-
- {%- else -%}
-
- {%- for project in sorted_projects -%}
- {% include projects.html %}
- {%- endfor %}
-
- {%- endif -%}
-{%- endif -%}
-
diff --git a/_pages/_teaching.md b/_pages/_teaching.md
deleted file mode 100644
index 1a6c5be..0000000
--- a/_pages/_teaching.md
+++ /dev/null
@@ -1,12 +0,0 @@
----
-layout: page
-permalink: /teaching/
-title: teaching
-description: Materials for courses you taught. Replace this text with your description.
-nav: true
-nav_order: 4
----
-
-For now, this page is assumed to be a static description of your courses. You can convert it to a collection similar to `_projects/` so that you can have a dedicated page for each course.
-
-Organize your courses by years, topics, or universities, however you like!
diff --git a/_pages/about.md b/_pages/about.md
deleted file mode 100644
index 7ff6791..0000000
--- a/_pages/about.md
+++ /dev/null
@@ -1,33 +0,0 @@
----
-layout: about
-title: About
-permalink: /
-subtitle: Welcome to HKU NLP group! We are a group of researchers working on natural language processing in the Department of Computer Science at the University of Hong Kong.
-logo: group_name.png
-research: wordcloud.png
-
-profile:
- align: left #right
- image: group_photo_2022.jpeg
- image_cicular: false # crops the image to make it circular
- image_caption: Group members in Hong Kong as of September 12, 2022 @ Tai Long Wan
-
-
-
-
-news: true # includes a list of news items
-selected_papers: false # includes a list of papers marked as "selected={true}"
-social: true # includes social icons at the bottom of the page
-
----
-The current focus of the NLP group in HKU includes:
-• Pretraining algorithms, including novel objectives, neural network architectures and distillation methods;
-• Semantic parsing, natural language interfaces, human-computer interaction (HCI);
-• Dialog systems, long text generation and automatic evaluation;
-• Machine Translation, especially in the settings of simultaneous and multimodal MT.
-
-## Overview
-This blog post introduces a new perspective to understanding the **Random Feature Attention (RFA)** mechanism. We show that **1)** the conventional softmax attention can be equivalently rewritten as an **expectation over RFAs**, and that **2)** RFA is in fact a **self-normalized importance sampler** to estimate conventional softmax attention. This new perspective grounds the heuristic RFA approximation and also sheds light on how to generalize further and improve RFAs. More details can be found in our ICML paper .
-
-## Attention
-The attention mechanism has become a ubiquitous building block in modern deep learning models and brought great success across various domains, including natural language processing (NLP), computer vision (CV), bioinformatics, reinforcement learning, etc.
-Attention mechanisms take three different kinds of inputs: a set of $N$ query vectors $\mbQ \in \R^{N \times D}$, $M$ key vectors $\mbK \in \R^{M \times D}$ and value vectors $\mbV \in \R^{M \times D}$. In this work we focus on self-attention, where $M=N$ and all of queries, keys and values are obtained by projecting tokens of the same input sequence.
-
-For each query $\mbq_n$, conventional softmax attention computes the following quantity, We omit the commonly used scaling factor $1 / \sqrt{d}$ for simplicity as it can be merged into the computation of queries or keys.
-\begin{equation}
- \mathsf{SoftmaxAttn}\left(\mbq_{n},\mbK,\mbV\right)\coloneqq\sum_{m=1}^M\frac{\exp\left(\mbq_{n}^\top \mbk_{m} \right)}{\sum_{m'=1}^M \exp\left(\mbq_{n}^\top \mbk_{m'} \right)} \mbv_{m}^{\top}.
-\end{equation}
-Intuitively, softmax attention first compares the query against each key and then computes the average over value vectors weighted by the normalized query-key similarities. It is effective in capturing long-term dependencies across sequence elements and producing contextualized representations; however, it suffers from **quadratic** time and memory complexity due to the explicit computation of all $NM$ query-key pairs.
-
-
-## Random Feature Attention
-To reduce the computational complexity of softmax attention, researchers proposed to use **random features (RF)** to linearize softmax attention. In particular, they make use of the following identity to rewrite the **exponential kernel** as an expectation,
-\begin{equation}
- \exp(\mbx^\top \mby) = \mathbb{E}_{\omega \sim \mathcal{N}(\omega;0,\mathbf{I})}\left[\xi(\mbx,\omega)^\top\xi(\mby, \omega)\right], \label{eqn:identity}
-\end{equation}
-where $\xi(\cdot, \cdot): \R^D \times \R^D \rightarrow \R^l$, $l\geq 1$, is a non-linear **randomized mapping** projecting the input vector to a $l$-dimensional vector via a randomly drawn $\omega \sim \mathcal{N}(\omega;0,\mathbf{I})$. There are several parameterization choices of $\xi(\mbx,\omega)$ for the identity to hold; in this work we focus on the positive type $\xi(\mbx,\omega) = \exp{\left(\omega^\top \mbx - \frac{1}{2}\norm{\mbx}^2\right)}$ as proposed in . A classical choice of randomized mappings is to let $\xi(\mbx,\omega) = \exp{\left( \frac{1}{2}\norm{\mbx}^2\right)}\left[\sin{\left(\omega^\top \mbx\right)},\cos{\left(\omega^\top \mbx\right)}\right]^\top$ . Besides, there are other advanced randomized mappings enjoying more appealing properties; see for a more in-depth study.
-
-
-To estimate the expectation in \eqref{eqn:identity}, one can draw multiple Monte Carlo samples from $\mathcal{N}(\omega;0,\mathbf{I})$ such that $\exp(\mbx^{\top} \mby) \approx \frac{1}{S}\sum_{s=1}^S \xi(\mbx,\omega_s)^{\top}\xi(\mby, \omega_s)$. By substituting such approximation into the softmax attention, we obtain random feature attention (RFA) (also called Performer),
-$$
-\begin{align}
- \frac{\sum_{m=1}^M\exp\left(\mbq_{n}^\top \mbk_{m} \right)\mbv_{m}^{\top}}{\sum_{m'=1}^M \exp\left(\mbq_{n}^\top \mbk_{m'} \right)}
- &\approx
- \frac{\sum_{m=1}^M \sum_{s=1}^S\xi(\mbq_n,\omega_s)^{\top}\xi(\mbk_m, \omega_s)\mbv_{m}^{\top}}{\sum_{m'=1}^M\sum_{s=1}^S \xi(\mbq_n,\omega_s)^{\top}\xi(\mbk_{m'}, \omega_s)} \notag \\
- &=\frac{ \sum_{s=1}^S\xi(\mbq_n,\omega_s)^{\top}\sum_{m=1}^M\xi(\mbk_m, \omega_s)\mbv_{m}^{\top}}{\sum_{s=1}^S \xi(\mbq_n,\omega_s)^{\top}\sum_{m'=1}^M\xi(\mbk_{m'}, \omega_s)} \label{eqn:rfa}\\
- &\coloneqq \mathsf{RFA}\left(\mbq_{n},\mbK,\mbV\right) \notag.
-\end{align}
-$$
-Thanks to the linearized formulation, one can first pre-compute the corresponding key-value statistics $\sum_{m=1}^M\xi(\mbk_{m},\omega_s)\mbv_{m}^{\top}$ and $\sum_{m=1}^M\xi(\mbk_{m},\omega_s)$ once, and then reuse them for each query. Consequently, it achieves linear complexity in both time and memory with respect to the sequence length.
-
Why it is called random feature attention?
- This is due to the fact that the sample average can be written as $\mbphi(\mbx,\mbw)^\top \mbphi(\mby,\mbw)$, where $\mbphi(\mbx,\mbw) \coloneqq 1/\sqrt{S}[\xi(\mbx,\omega_1), \dots, \xi(\mbx, \omega_S)]^\top \in \R^{lS}$. The $\mbphi(\cdot,\cdot)$ can be considered as a feature map transforming the input vector to a new vector representation; as a result, $\mbphi(\cdot,\cdot)$ are conveniently referred to as random features.
-
-## The Biasedness of RFA
-RFA suffers from significant performance degradation across various tasks, as observed in many previous studies (e.g., see [here](#comp-tb)). Although computationally efficient, the reduced modeling capacity of RFA greatly limits its practical usage. Researchers try to improve the performance from different perspectives, such as
-- developing better random feature types ,
-- connecting it with fast weight programmers ,
-- incorporating relative information , and so on.
-
-In this work, we explore an orthogonal axis by re-examining the estimation bias of RFA. Our key observation is that RFA is a heuristic approximation to the **whole** softmax attention. Although the estimation of individual **exponential kernels** is unbiased, such estimation of a ratio of exponential kernels is not unbiased anymore. This is due to the non-linearity of ratios,
-
-Such bias not only incurs a potentially large approximation gap between RFA and softmax attention but also bottlenecks the effectiveness of unbiased kernel estimation.
-
-
-
- Our work aims to address the following research question:
- Given that we already know how to unbiasedly estimate exponential kernels, how do we construct an unbiased estimator for the whole softmax attention?
-
-
-
-## Randomized Attention: An Unbiased Estimator for Softmax Attention
-
-We answer this question in the affirmative and prove that softmax attention can be rewritten as an expectation of simple RFAs,
-$$
-\begin{equation}
-\mathsf{SoftmaxAttn}(\mbq_n, \mbK,\mbV) = \sum_{m} \frac{\exp \left(\mathbf{q}_{n}^\top \mathbf{k}_{m} \right)}{\sum_{m'} \exp \left(\mathbf{q}_{n}^\top \mathbf{k}_{m'} \right)} \mathbf{v}_{m}^{\top} = \mathbb{E}_{p_n(\omega)}\left[f_n(\omega)\right],\label{eqn:softmax_as_expectation}
-\end{equation}
-$$
-where
-$$
-\begin{align}
-p_n(\omega) &= \sum_{m=1}^M \frac{\exp\left( \mbq_n^\top\mbk_m \right)}{\sum_{m'=1}^M\exp\left( \mbq_n^\top\mbk_{m'} \right)} \mathcal{N}(\omega; \mbq_n + \mbk_m, \mathbf{I}), \label{eqn:ra-density} \\
-f_n(\omega) &= \frac{\xi(\mbq_n,\omega)^\top \sum_{m=1}^M \xi(\mbk_m, \omega) \mbv_{m}^{\top}}{\xi(\mbq_n,\omega)^\top \sum_{m'=1}^M \xi(\mbk_{m'}, \omega)} \label{eqn:ra-function}.
-\end{align}
-$$
-Intuitively,
-
-
$p_n(\omega)$ is a Gaussian mixture distribution whose component weights are exactly attention scores, while
-
$f_n(\omega)$ is an RFA, except that the linearized similarity between queries and keys is computed via individual randomized mappings (or equivalently, one-dimensional random features).
-
-
-
-Notably, our result can be viewed as a neat generalization of random feature approximation, which exhibits a high degree of symmetry:
-$$
-\begin{align}
- \exp(\mbq_n^\top \mbk_m)\mbv_m^\top &= \mathbb{E}_{q(\omega)}\left[\xi(\mbq_n,\omega)^\top\xi(\mbk_m, \omega)\mbv_m^\top\right] \notag \\
- \sum_{m} \frac{\exp \left(\mathbf{q}_{n}^\top \mathbf{k}_{m} \right)}{\sum_{m'} \exp \left(\mathbf{q}_{n}^\top \mathbf{k}_{m'} \right)} \mathbf{v}_{m}^{\top} &= \mathbb{E}_{p_n(\omega)}\left[\frac{\xi(\mbq_n,\omega)^\top \sum_{m=1}^M \xi(\mbk_m, \omega) \mbv_{m}^{\top}}{\xi(\mbq_n,\omega)^\top \sum_{m'=1}^M \xi(\mbk_{m'}, \omega)}\right].\notag
-\end{align}
-$$
-Another implication is that we can construct a Monte Carlo estimator to approximate softmax attention in an unbiased way. By drawing $\omega_n \sim p_n(\omega)$, we obtain
-$$
-\begin{align}
- \mathsf{SoftmaxAttn}(\mbq_n, \mbK,\mbV)
- &\approx \frac{\xi(\mbq_n,\omega_n)^\top \sum_{m=1}^M\xi(\mbk_m, \omega_n) \mbv_{m}^{\top}}{ \xi(\mbq_n,\omega_n)^\top \sum_{m'=1}^M\xi(\mbk_{m'}, \omega_n)} \label{eqn:ra}\\
- &\coloneqq \mathsf{RA}\left(\mbq_{n},\mbK,\mbV\right) \notag
-\end{align}
-$$
-We name such estimator Randomized Attention (RA) since it computes similarity scores with individual randomized mappings (instead of concatenated features). To the best of our knowledge, this is the first result that generalizes the unbiased kernel estimation to unbiased attention estimation.
-
-
-
-
Remark:
- The proof of \eqref{eqn:softmax_as_expectation} is done by first reverse-engineering the formulation of RFA, equating it with self-normalized importance sampling (see below) and then completing the square of Gaussians to derive the density $p_n(\omega)$. The function $f_n(\omega)$ can be solved by substituting the density $p_n(\omega)$ into the equation. See the paper for a detailed proof.
-
-
-## RFA as a Self-normalized Importance Sampler
-The analysis above further reveals that RFA is a specific **self-normalized importance sampling** estimator to softmax attention.
-
-### Self-normalized Importance Sampling (SNIS)
-
-Importance sampling (IS) is a general approach to approximating expectation $\mathbb{E}_{p_n(\omega)}\left[f_n(\omega)\right]$ when it is difficult to directly draw samples from $p_n(\omega)$. In importance sampling, we use a proposal distribution $q(\omega)$ to draw samples and estimate the quantity as
-$$
-\mathbb{E}_{p_n(\omega)}\left[f_n(\omega)\right] = \mathbb{E}_{\omega \sim q(\omega)}\left[\frac{p_n(\omega)}{q(\omega)}f_n(\omega)\right] \approx \frac{1}{S} \sum_{s=1}^S \frac{p_n(\omega)}{q(\omega)} f_n(\omega_s),
-$$
-where $\omega_1, \dots, \omega_S \sim q(\omega)$. The self-normalized importance sampling (SNIS) is defined as
-$$
-\begin{equation}
- \mathbb{E}_{p_n(\omega)}\left[f_n(\omega)\right] \approx \frac{\sum_{s=1}^S\frac{p_n(\omega_s)}{q(\omega_s)}f(\omega_s)}{\sum_{s=1}^S\frac{p_n(\omega_s)}{q(\omega_s)}} = \sum_{s=1}^S\frac{\frac{p_n(\omega_s)}{q(\omega_s)}}{\sum_{s=1}^S\frac{p_n(\omega_s)}{q(\omega_s)}}f(\omega_s). \label{eqn:snis}
-\end{equation}
-$$
-The name self-normalized comes from the fact that the importance weights $p_n(\omega)/q(\omega)$ are explicitly normalized and sum to 1.
-
-
-
-
-
-### RFA as SNIS
-
Our key finding here is that the formulation of RFA can be exactly derived from SNIS. Supposing $p_n(\omega)$ and $f_n(\omega)$ are given in \eqref{eqn:ra-density} and \eqref{eqn:ra-function} respectively, and $q(\omega) = \mathcal{N}(\omega;0,\mathbf{I})$, we have
-$$
-\begin{align}
-\mathsf{RFA}\left(\mbq_{n},\mbK,\mbV\right)
-&= \frac{\sum_{s=1}^S\textcolor{strings}{\xi(\mbq_n,\omega_s)^{\top}\sum_{m=1}^M\xi(\mbk_m, \omega_s)\mbv_{m}^{\top}}}{\sum_{s=1}^S \textcolor{keywords}{\xi(\mbq_n,\omega_s)^{\top}\sum_{m'=1}^M\xi(\mbk_{m'}, \omega_s)}} = \frac{ \sum_{s=1}^S\textcolor{strings}{\frac{p_n(\omega_s)}{q(\omega_s)} f(\omega_s)}}{ \sum_{s=1}^S\textcolor{keywords}{\frac{p_n(\omega_s)}{q(\omega_s)}}}. \label{eqn:rfa-as-snis}
-\end{align}
-$$
-This formulation provides a new understanding of RFA: it is just a specific instantiation of SNIS estimators for softmax attention, whose proposal distribution $q(\omega)$ is chosen to be standard Gaussian. This reveals one of the possible reasons why RFA does not work well in practice: The plain standard Gaussian proposal in RFA is far away from the true Gaussian mixture (as in RA), which might lead to a large approximation gap. More importantly, this view implies that we can generalize and extend RFA by using other proposal distributions or adopting other estimating schemes!
-
-## LARA: Generalizing Both RA and RFA
-So far, we have two types of estimators available for approximating softmax attention: unbiased RA and biased RFA. Besides the theoretical biasedness, how do they differ in terms of practical modeling behavior? We list a comprehensive comparison to better illustrate their main differences.
-
-- In terms of **expressiveness**:
- - ✔️ RA directly draws from the true distribution $p_n(\omega)$ \eqref{eqn:ra-density}. This makes the mechanism **adaptive** and **query-specific**, since the sampling distribution depends on the query vector. As a result, RA can **specialize** to each query and process the whole sequence at a finer-grained level.
- - ❌ RFA suffers from limited capacity, since there is a large discrepancy between the used proposal $q(\omega) = \mathcal{N}(\omega;0,\mathbf{I})$ and the true distribution $p_n(\omega)$; furthermore, $q(\omega)$ captures no contextual information, leading to low sample efficiency.
-- In terms of **efficiency**:
- - ❌ RA suffers from **quadratic** computational costs. Since the sampling distribution is distinct for each query, we have to draw at least $\mathcal{O}(N)$ samples for all queries; at the same time, we needs to sum over all $M$ key-value pairs at each sampled $\omega_n$ (see \eqref{eqn:ra}), which leads to $\mathcal{O}(NM)$ complexity overall, as is the case for softmax attention;
- - ✔️ RFA uses SNIS to avoid computing or sampling from $p_n(\omega)$, which require quadratic-time computation. Drawing from a simpler proposal distribution allows samples to be shared across queries and still resulting in a $\mathcal{O}(M)$ complexity.
-
-
- {% include figure.html path="assets/img/2022-07-12-lara-imgs/lara-as-combining-ra-and-rfa.png" class="img-fluid rounded z-depth-0" zoomable=true caption="A diagram of LARA that combines the strengths of both approaches." %}
-
-
-
-
-
-
-Motivated by the comparison, we propose LineAr-time Randomized Attention (LARA) that attempts to get the best of both worlds, combining both the efficiency of RFA and the expressiveness of RA.
-- On the one hand, LARA inherits the **SNIS** formulation as in RFA to achieve **linear complexity**;
-- On the other hand, its expressiveness is significantly improved towards RA via the following modifications:
- - The proposal distribution is data-dependent so that LARA can be **adaptive** with respect to the contextual information;
- - LARA adopts **multiple** proposal distributions, which can be combined in a **query-specific** manner.
-
-
-LARA takes the following form
-$$
-\begin{align}
-\mathsf{LARA}\left(\mbq_{n},\mbK,\mbV\right) &= \frac{\sum_{c=1}^C\alpha_{nc}(\omega_c)\frac{p_n(\omega_c)}{q_c(\omega_c)} f_n(\omega_c)}{\sum_{c=1}^C\alpha_{nc}(\omega_c)\frac{p_n(\omega_c)}{q_c(\omega_c)}}.\label{eqn:lara}
-\end{align}
-$$
-Here,
-
-
$q_1(\omega), \dots, q_C(\omega)$ are $C$ different proposals that depend on different query and key information;
-
$\{\alpha_{nc}(\cdot)\}_{c=1}^C$ is a set of weights that controls the contribution of different proposals, which is exclusive to each query and satisfies $\sum_{c=1}^C \alpha_{nc}(\omega) = 1$;
-
as long as $C$ is smaller than $N$, the computational complexity of LARA is linear, that is, $\mathcal{O}(CN+CM)$.
-
-
-
-
Remark:
- We provide a detailed discussion about the parameterization of our proposal distributions in Appendix G.3 of our paper. To summarize, we find the key is to let different proposals depend on different sets of query information so that they could be as query-specific as possible. A good default is to divide the whole sequence into $C$ chunks, compute the mean of queries $\{\widetilde{\mbq}_c\}_{c=1}^C$ and keys $\{\widetilde{\mbk}_c\}_{c=1}^C$ within the same chunk, and set $q_c(\omega) = \mcN(\omega;\widetilde{\mbq}_c + \widetilde{\mbk}_c, \mathbf{I})$. We find this choice works well across various benchmarks.
-
-
-### A Unified View of LARA, RA, and RFA
-
LARA can be equivalently written in a similar way to RA and RFA. We spell it out here to see a systematic comparison among RA, LARA, and RFA:
-$$
-\begin{align}
-\mathsf{RA}\left(\mbq_{n},\mbK,\mbV\right) &= \frac{\xi(\mbq_n,\omega_n)^\top \sum_{m=1}^M\xi(\mbk_m, \omega_n) \mbv_{m}^{\top}}{ \xi(\mbq_n,\omega_n)^\top \sum_{m'=1}^M\xi(\mbk_{m'}, \omega_n)}, &&\textcolor{keywords}{\omega_n \sim p_n(\omega)}\notag\\
-\mathsf{LARA}\left(\mbq_{n},\mbK,\mbV\right) &= \frac{\sum_{c=1}^C \textcolor{strings}{\alpha'_{nc}(\omega_c)} \xi(\mbq_n,\omega_c)^\top \sum_{m=1}^M\xi(\mbk_m, \omega_c) \mbv_{m}^{\top}}{\sum_{c=1}^C \textcolor{strings}{\alpha'_{nc}(\omega_c)} \xi(\mbq_n,\omega_c)^\top \sum_{m=1}^M \xi(\mbk_{m}, \omega_c)}, &&\textcolor{keywords}{\omega_c \sim q_c(\omega)}\notag\\
-\mathsf{RFA}\left(\mbq_{n},\mbK,\mbV\right) &= \frac{ \sum_{s=1}^S\xi(\mbq_n,\omega_s)^{\top}\sum_{m=1}^M\xi(\mbk_m, \omega_s)\mbv_{m}^{\top}}{\sum_{s=1}^S \xi(\mbq_n,\omega_s)^{\top}\sum_{m'=1}^M\xi(\mbk_{m'}, \omega_s)}, &&\textcolor{keywords}{\omega_1,\dots,\omega_S \sim \mcN(\omega;0, \mathbf{I})} \notag
-\end{align}
-$$
-where we denote $\alpha'_{nc}(\omega_c) \coloneqq \alpha_{nc}(\omega_c)\mcN(\omega_c;0, \mathbf{I})/q_c(\omega_c)$ to simplify the notation. Note that their major difference lies in the choice of sampling distributions.
-
-LARA is not designed to be a simple interpolation between RA and RFA; instead, it is a generalized estimation framework that includes both RA and RFA as its special cases. To see this,
-
-
LARA is equivalent to RA if $C = N$, $q_n(\omega) = p_n(\omega)$ and $\alpha_{nc}(\omega) = \delta_{nc}$ (that is, $\alpha_{nc}(\omega) = 1$ if $n = c$ and $0$ otherwise);
-
LARA is equivalent to RFA if $q_c(\omega) = \mcN(\omega;0,\mathbf{I})$ and $\alpha_{nc}(\cdot)$ is constant for all $c = 1,\dots,C$ and $n = 1,\dots,N$.
-
-With general proposals and weighting functions, LARA approximates softmax attention in a query-specific manner as in RA while achieving linear complexity as in RFA, effectively combining the advantages of both estimators. It is also easy to implement with a couple of lines of code.
-
-
-
-
-
-
-## Experimental Results
-
-To demonstrate the effectiveness of our approach, we first visualize the approximation error of LARA to the true softmax attention outputs. Our unbiased estimate, RA, achieves the lowest MSE among these three methods and gets very close to the true softmax attention; RFA (Performer) soon plateaus at large approximation error and does not improve even with more samples, possibly due to low sample efficiency. On the other hand, LARA exhibits much lower MSE than RFA and the approximation error continually decreases as the number of proposals increases.
-
-
- {% include figure.html path="assets/img/2022-07-12-lara-imgs/approx_error_deit_plot.png" class="img-fluid rounded z-depth-1" zoomable=true caption="Mean Squared Error (MSE) between the true softmax attention and different RF approximations under different numbers of
-samples (lower is better), which are evaluated on DeiT." %}
-
-
-
-We further verify the improved performance of LARA by conducting experiments across a wide range of data modalities, including images, videos, natural language texts, and a long-sequence benchmark. From the Table below, we observe that
-- RA performs competitively with conventional softmax attention, and
-- LARA greatly improves the performance of RFA and significantly reduces the performance gap between RFA and softmax attention.
-
-| Model | Complexity | Image Classification on ImageNet (🠅) | Video Recognition on SSv2 (🠅) | Machine Translation on WMT (🠅) | Long Range Arena suite (🠅) |
-|:---|:---:|:---:|:---:|:---:|:---:|
-| Softmax | $\mathcal{O}(N^2)$ | 79.9 | 66.5 | 27.5 | 59.08 |
-| RA | $\mathcal{O}(N^2)$ | 80.0 | 64.9 | 27.8 | 59.30 |
-| RFA | $\mathcal{O}(N)$ | 74.3 | 53.1 | 23.7 | 57.63 |
-| LARA | $\mathcal{O}(N)$ | **79.5** | **63.7** | **27.0** | **59.12** |
-
-LARA also enjoys much better scalability and could achieve SOTA results for image classification when applying it to advanced transformer architectures.
-
-
- {% include figure.html path="assets/img/2022-07-12-lara-imgs/imagenet-sota-lara.png" class="img-fluid rounded z-depth-1" zoomable=true caption="SOTA results on ImageNet-1k across various model architectures." %}
-
-
-
-
-Finally, we evaluate the empirical efficiency of different attention methods. We note that RA runs almost twice slower than softmax attention, while its linear variant LARA runs much faster and brings marginal computational overheads compared to RFA.
-
-
- {% include figure.html path="assets/img/2022-07-12-lara-imgs/mem-time-comparison.png" class="img-fluid rounded z-depth-1" zoomable=true caption="Empirical memory consumption (left) and running time (right) of different attention mechanisms under different sequence
-lengths. Metrics are measured relative to the true softmax attention." %}
-
-
-
-## Conclusion and Limitation
-In this work, we generalize the random feature attention (RFA) algorithm in two aspects:
-- On the one hand, we generalize the unbiased **kernel** estimation (RFA) to unbiased **attention** estimation (RA);
-- On the other hand, we uncover the self-normalized importance sampling nature of RFA and propose a general framework **LARA** that includes both RA and RFA as its special cases.
-
-LARA greatly improves the performance of RFA while also maintaining its efficiency of RFA. Our framework provides a novel perspective for understanding and improving RF-based attention approximation, which is also orthogonal to most previous work. At the same time, there are several limitations of our approach:
-- The choice of proposal distributions is sub-optimal. Although our modeling strategy enables a query-specific treatment as desired in RA, it needs to evaluate all the $C$ samples for each key-value pair, incurring a potential waste of computing. Future work might include more sophisticated implementations that use variable sample sets for different key-value pairs.
-- Currently, LARA is still a biased (but consistent) estimator for softmax attention. It would be interesting to adopt advanced sampling strategies (e.g., Markov chain Monte Carlo (MCMC), etc.) to derive lower-bias lower-variance estimators.
-- The proposal distribution should be carefully designed when dealing with auto-regressive modeling since it is non-trivial to apply the causal masking mechanism to LARA as in conventional softmax attention. Adapting the importance sampling formulation of LARA to support arbitrary auto-regressive modeling is also an interesting research direction.
-
-
-
diff --git a/_posts/_2015-03-15-formatting-and-links.md b/_posts/_2015-03-15-formatting-and-links.md
deleted file mode 100644
index 6a2d63e..0000000
--- a/_posts/_2015-03-15-formatting-and-links.md
+++ /dev/null
@@ -1,359 +0,0 @@
----
-layout: post
-title: a post with formatting and links
-date: 2015-03-15 16:40:16
-description: march & april, looking forward to summer
-tags: formatting links
-categories: sample-posts
-# redirect: /assets/pdf/example_pdf.pdf
----
-Jean shorts raw denim Vice normcore, art party High Life PBR skateboard stumptown vinyl kitsch. Four loko meh 8-bit, tousled banh mi tilde forage Schlitz dreamcatcher twee 3 wolf moon. Chambray asymmetrical paleo salvia, sartorial umami four loko master cleanse drinking vinegar brunch. Pinterest DIY authentic Schlitz, hoodie Intelligentsia butcher trust fund brunch shabby chic Kickstarter forage flexitarian. Direct trade cold-pressed meggings stumptown plaid, pop-up taxidermy. Hoodie XOXO fingerstache scenester Echo Park. Plaid ugh Wes Anderson, freegan pug selvage fanny pack leggings pickled food truck DIY irony Banksy.
-
-#### Hipster list
-
-
brunch
-
fixie
-
raybans
-
messenger bag
-
-
-Hoodie Thundercats retro, tote bag 8-bit Godard craft beer gastropub. Truffaut Tumblr taxidermy, raw denim Kickstarter sartorial dreamcatcher. Quinoa chambray slow-carb salvia readymade, bicycle rights 90's yr typewriter selfies letterpress cardigan vegan.
-
-
-
-Pug heirloom High Life vinyl swag, single-origin coffee four dollar toast taxidermy reprehenderit fap distillery master cleanse locavore. Est anim sapiente leggings Brooklyn ea. Thundercats locavore excepteur veniam eiusmod. Raw denim Truffaut Schlitz, migas sapiente Portland VHS twee Bushwick Marfa typewriter retro id keytar.
-
-
- We do not grow absolutely, chronologically. We grow sometimes in one dimension, and not in another, unevenly. We grow partially. We are relative. We are mature in one realm, childish in another.
- —Anais Nin
-
-
-Fap aliqua qui, scenester pug Echo Park polaroid irony shabby chic ex cardigan church-key Odd Future accusamus. Blog stumptown sartorial squid, gastropub duis aesthetic Truffaut vero. Pinterest tilde twee, odio mumblecore jean shorts lumbersexual.
-
-
-
-# Tweet
-An example of displaying a tweet:
-{% twitter https://twitter.com/rubygems/status/518821243320287232 %}
-
-# Timeline
-An example of pulling from a timeline:
-{% twitter https://twitter.com/jekyllrb maxwidth=500 limit=3 %}
-
-# Additional Details
-For more details on using the plugin visit: [jekyll-twitter-plugin](https://github.com/rob-murray/jekyll-twitter-plugin)
-
-
-This theme supports generating various diagrams from a text description using [jekyll-diagrams](https://github.com/zhustec/jekyll-diagrams){:target="\_blank"} plugin.
-Below, we generate a few examples of such diagrams using languages such as [mermaid](https://mermaid-js.github.io/mermaid/){:target="\_blank"}, [plantuml](https://plantuml.com/){:target="\_blank"}, [vega-lite](https://vega.github.io/vega-lite/){:target="\_blank"}, etc.
-
-**Note:** different diagram-generation packages require external dependencies to be installed on your machine.
-Also, be mindful of that because of diagram generation the fist time you build your Jekyll website after adding new diagrams will be SLOW.
-For any other details, please refer to [jekyll-diagrams](https://github.com/zhustec/jekyll-diagrams){:target="\_blank"} README.
-
-
-## Mermaid
-
-Install mermaid using `node.js` package manager `npm` by running the following command:
-```bash
-npm install -g mermaid.cli
-```
-
-The diagram below was generated by the following code:
-
-{% raw %}
-```
-{% mermaid %}
-sequenceDiagram
- participant John
- participant Alice
- Alice->>John: Hello John, how are you?
- John-->>Alice: Great!
-{% endmermaid %}
-```
-{% endraw %}
-
-{% mermaid %}
-sequenceDiagram
- participant John
- participant Alice
- Alice->>John: Hello John, how are you?
- John-->>Alice: Great!
-{% endmermaid %}
-
-
-
-A sample blog page that demonstrates the accessing of github meta data.
-
-## What does Github-MetaData do?
-* Propagates the site.github namespace with repository metadata
-* Setting site variables :
- * site.title
- * site.description
- * site.url
- * site.baseurl
-* Accessing the metadata - duh.
-* Generating edittable links.
-
-## Additional Reading
-* If you're recieving incorrect/missing data, you may need to perform a Github API authentication.
-* Go through this README for more details on the topic.
-* This page highlights all the feilds you can access with github-metadata.
-
-
-## Example MetaData
-* Host Name : {{ site.github.hostname }}
-* URL : {{ site.github.url }}
-* BaseURL : {{ site.github.baseurl }}
-* Archived : {{ site.github.archived}}
-* Contributors :
-{% for contributor in site.github.contributors %}
- * {{ contributor.login }}
-{% endfor %}
-
-
-This theme supports rendering beautiful math in inline and display modes using [MathJax 3](https://www.mathjax.org/) engine. You just need to surround your math expression with `$$`, like `$$ E = mc^2 $$`. If you leave it inside a paragraph, it will produce an inline expression, just like $$ E = mc^2 $$.
-
-To use display mode, again surround your expression with `$$` and place it as a separate paragraph. Here is an example:
-
-$$
-\sum_{k=1}^\infty |\langle x, e_k \rangle|^2 \leq \|x\|^2
-$$
-
-You can also use `\begin{equation}...\end{equation}` instead of `$$` for display mode math.
-MathJax will automatically number equations:
-
-\begin{equation}
-\label{eq:cauchy-schwarz}
-\left( \sum_{k=1}^n a_k b_k \right)^2 \leq \left( \sum_{k=1}^n a_k^2 \right) \left( \sum_{k=1}^n b_k^2 \right)
-\end{equation}
-
-and by adding `\label{...}` inside the equation environment, we can now refer to the equation using `\eqref`.
-
-Note that MathJax 3 is [a major re-write of MathJax](https://docs.mathjax.org/en/latest/upgrading/whats-new-3.0.html) that brought a significant improvement to the loading and rendering speed, which is now [on par with KaTeX](http://www.intmath.com/cg5/katex-mathjax-comparison.php).
-
-This is an example post with image galleries.
-
-
-
- {% include figure.html path="assets/img/9.jpg" class="img-fluid rounded z-depth-1" %}
-
-
- {% include figure.html path="assets/img/7.jpg" class="img-fluid rounded z-depth-1" %}
-
-
-
- A simple, elegant caption looks good between image rows, after each row, or doesn't have to be there at all.
-
-
-Images can be made zoomable.
-Simply add `data-zoomable` to `` tags that you want to make zoomable.
-
-
-
-This theme implements a built-in Jekyll feature, the use of Rouge, for syntax highlighting.
-It supports more than 100 languages.
-This example is in C++.
-All you have to do is wrap your code in a liquid tag:
-
-{% raw %}
-{% highlight c++ linenos %} code code code {% endhighlight %}
-{% endraw %}
-
-The keyword `linenos` triggers display of line numbers.
-Produces something like this:
-
-{% highlight c++ linenos %}
-
-int main(int argc, char const \*argv[])
-{
- string myString;
-
- cout << "input a string: ";
- getline(cin, myString);
- int length = myString.length();
-
- char charArray = new char * [length];
-
- charArray = myString;
- for(int i = 0; i < length; ++i){
- cout << charArray[i] << " ";
- }
-
- return 0;
-}
-
-{% endhighlight %}
-
-
-## Citations
-
-Citations are then used in the article body with the `` tag.
-The key attribute is a reference to the id provided in the bibliography.
-The key attribute can take multiple ids, separated by commas.
-
-The citation is presented inline like this: (a number that displays more information on hover).
-If you have an appendix, a bibliography is automatically created and populated in it.
-
-## Code Blocks
-
-Syntax highlighting is provided within `` tags.
-An example of inline code snippets: `let x = 10;`.
-For larger blocks of code, add a `block` attribute:
-
-
- var x = 25;
- function(x) {
- return x * x;
- }
-
-
-**Note:** `` blocks do not look well in the dark mode.
-You can always use the default code-highlight using the `highlight` liquid tag:
-
-{% highlight javascript %}
-var x = 25;
-function(x) {
- return x * x;
-}
-{% endhighlight %}
-
-***
-
-
-
-***
-
-## Other Typography?
-
-Emphasis, aka italics, with *asterisks* (`*asterisks*`) or _underscores_ (`_underscores_`).
-
-Strong emphasis, aka bold, with **asterisks** or __underscores__.
-
-Combined emphasis with **asterisks and _underscores_**.
-
-Strikethrough uses two tildes. ~~Scratch this.~~
-
-1. First ordered list item
-2. Another item
-⋅⋅* Unordered sub-list.
-1. Actual numbers don't matter, just that it's a number
-⋅⋅1. Ordered sub-list
-4. And another item.
-
-⋅⋅⋅You can have properly indented paragraphs within list items. Notice the blank line above, and the leading spaces (at least one, but we'll use three here to also align the raw Markdown).
-
-⋅⋅⋅To have a line break without a paragraph, you will need to use two trailing spaces.⋅⋅
-⋅⋅⋅Note that this line is separate, but within the same paragraph.⋅⋅
-⋅⋅⋅(This is contrary to the typical GFM line break behaviour, where trailing spaces are not required.)
-
-* Unordered list can use asterisks
-- Or minuses
-+ Or pluses
-
-[I'm an inline-style link](https://www.google.com)
-
-[I'm an inline-style link with title](https://www.google.com "Google's Homepage")
-
-[I'm a reference-style link][Arbitrary case-insensitive reference text]
-
-[I'm a relative reference to a repository file](../blob/master/LICENSE)
-
-[You can use numbers for reference-style link definitions][1]
-
-Or leave it empty and use the [link text itself].
-
-URLs and URLs in angle brackets will automatically get turned into links.
-http://www.example.com or and sometimes
-example.com (but not on Github, for example).
-
-Some text to show that the reference links can follow later.
-
-[arbitrary case-insensitive reference text]: https://www.mozilla.org
-[1]: http://slashdot.org
-[link text itself]: http://www.reddit.com
-
-Here's our logo (hover to see the title text):
-
-Inline-style:
-
-
-Reference-style:
-![alt text][logo]
-
-[logo]: https://github.com/adam-p/markdown-here/raw/master/src/common/images/icon48.png "Logo Title Text 2"
-
-Inline `code` has `back-ticks around` it.
-
-```javascript
-var s = "JavaScript syntax highlighting";
-alert(s);
-```
-
-```python
-s = "Python syntax highlighting"
-print s
-```
\ No newline at end of file
diff --git a/_projects/1_project.md b/_projects/1_project.md
deleted file mode 100644
index 5a95b9a..0000000
--- a/_projects/1_project.md
+++ /dev/null
@@ -1,80 +0,0 @@
----
-layout: page
-title: project 1
-description: a project with a background image
-img: assets/img/12.jpg
-importance: 1
-category: work
----
-
-Every project has a beautiful feature showcase page.
-It's easy to include images in a flexible 3-column grid format.
-Make your photos 1/3, 2/3, or full width.
-
-To give your project a background in the portfolio page, just add the img tag to the front matter like so:
-
- ---
- layout: page
- title: project
- description: a project with a background image
- img: /assets/img/12.jpg
- ---
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
-
- This image can also have a caption. It's like magic.
-
-
-You can also put regular text between your rows of images.
-Say you wanted to write a little bit about your project before you posted the rest of the images.
-You describe how you toiled, sweated, *bled* for your project, and then... you reveal it's glory in the next row of images.
-
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these.
-
-
-
-The code is simple.
-Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system).
-To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes.
-Here's the code for the last row of images above:
-
-{% raw %}
-```html
-
-```
-{% endraw %}
diff --git a/_projects/2_project.md b/_projects/2_project.md
deleted file mode 100644
index 4a26614..0000000
--- a/_projects/2_project.md
+++ /dev/null
@@ -1,80 +0,0 @@
----
-layout: page
-title: project 2
-description: a project with a background image
-img: assets/img/3.jpg
-importance: 2
-category: work
----
-
-Every project has a beautiful feature showcase page.
-It's easy to include images in a flexible 3-column grid format.
-Make your photos 1/3, 2/3, or full width.
-
-To give your project a background in the portfolio page, just add the img tag to the front matter like so:
-
- ---
- layout: page
- title: project
- description: a project with a background image
- img: /assets/img/12.jpg
- ---
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
-
- This image can also have a caption. It's like magic.
-
-
-You can also put regular text between your rows of images.
-Say you wanted to write a little bit about your project before you posted the rest of the images.
-You describe how you toiled, sweated, *bled* for your project, and then... you reveal it's glory in the next row of images.
-
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these.
-
-
-
-The code is simple.
-Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system).
-To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes.
-Here's the code for the last row of images above:
-
-{% raw %}
-```html
-
-```
-{% endraw %}
diff --git a/_projects/3_project.md b/_projects/3_project.md
deleted file mode 100644
index 56968db..0000000
--- a/_projects/3_project.md
+++ /dev/null
@@ -1,81 +0,0 @@
----
-layout: page
-title: project 3
-description: a project that redirects to another website
-img: assets/img/7.jpg
-redirect: https://unsplash.com
-importance: 3
-category: work
----
-
-Every project has a beautiful feature showcase page.
-It's easy to include images in a flexible 3-column grid format.
-Make your photos 1/3, 2/3, or full width.
-
-To give your project a background in the portfolio page, just add the img tag to the front matter like so:
-
- ---
- layout: page
- title: project
- description: a project with a background image
- img: /assets/img/12.jpg
- ---
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
-
- This image can also have a caption. It's like magic.
-
-
-You can also put regular text between your rows of images.
-Say you wanted to write a little bit about your project before you posted the rest of the images.
-You describe how you toiled, sweated, *bled* for your project, and then... you reveal it's glory in the next row of images.
-
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these.
-
-
-
-The code is simple.
-Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system).
-To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes.
-Here's the code for the last row of images above:
-
-{% raw %}
-```html
-
-```
-{% endraw %}
diff --git a/_projects/4_project.md b/_projects/4_project.md
deleted file mode 100644
index 96eeb0e..0000000
--- a/_projects/4_project.md
+++ /dev/null
@@ -1,80 +0,0 @@
----
-layout: page
-title: project 4
-description: another without an image
-img:
-importance: 3
-category: fun
----
-
-Every project has a beautiful feature showcase page.
-It's easy to include images in a flexible 3-column grid format.
-Make your photos 1/3, 2/3, or full width.
-
-To give your project a background in the portfolio page, just add the img tag to the front matter like so:
-
- ---
- layout: page
- title: project
- description: a project with a background image
- img: /assets/img/12.jpg
- ---
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
-
- This image can also have a caption. It's like magic.
-
-
-You can also put regular text between your rows of images.
-Say you wanted to write a little bit about your project before you posted the rest of the images.
-You describe how you toiled, sweated, *bled* for your project, and then... you reveal it's glory in the next row of images.
-
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these.
-
-
-
-The code is simple.
-Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system).
-To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes.
-Here's the code for the last row of images above:
-
-{% raw %}
-```html
-
-```
-{% endraw %}
diff --git a/_projects/5_project.md b/_projects/5_project.md
deleted file mode 100644
index aa48810..0000000
--- a/_projects/5_project.md
+++ /dev/null
@@ -1,80 +0,0 @@
----
-layout: page
-title: project 5
-description: a project with a background image
-img: assets/img/1.jpg
-importance: 3
-category: fun
----
-
-Every project has a beautiful feature showcase page.
-It's easy to include images in a flexible 3-column grid format.
-Make your photos 1/3, 2/3, or full width.
-
-To give your project a background in the portfolio page, just add the img tag to the front matter like so:
-
- ---
- layout: page
- title: project
- description: a project with a background image
- img: /assets/img/12.jpg
- ---
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
-
- This image can also have a caption. It's like magic.
-
-
-You can also put regular text between your rows of images.
-Say you wanted to write a little bit about your project before you posted the rest of the images.
-You describe how you toiled, sweated, *bled* for your project, and then... you reveal it's glory in the next row of images.
-
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these.
-
-
-
-The code is simple.
-Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system).
-To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes.
-Here's the code for the last row of images above:
-
-{% raw %}
-```html
-
-```
-{% endraw %}
diff --git a/_projects/6_project.md b/_projects/6_project.md
deleted file mode 100644
index 3d8d04d..0000000
--- a/_projects/6_project.md
+++ /dev/null
@@ -1,80 +0,0 @@
----
-layout: page
-title: project 6
-description: a project with no image
-img:
-importance: 4
-category: fun
----
-
-Every project has a beautiful feature showcase page.
-It's easy to include images in a flexible 3-column grid format.
-Make your photos 1/3, 2/3, or full width.
-
-To give your project a background in the portfolio page, just add the img tag to the front matter like so:
-
- ---
- layout: page
- title: project
- description: a project with a background image
- img: /assets/img/12.jpg
- ---
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
-
- This image can also have a caption. It's like magic.
-
-
-You can also put regular text between your rows of images.
-Say you wanted to write a little bit about your project before you posted the rest of the images.
-You describe how you toiled, sweated, *bled* for your project, and then... you reveal it's glory in the next row of images.
-
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these.
-
-
-
-The code is simple.
-Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system).
-To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes.
-Here's the code for the last row of images above:
-
-{% raw %}
-```html
-
- {%- endif -%}
- {%- elsif entry.abbr -%}
- {%- if site.data.venues[entry.abbr] -%}
- {{entry.abbr}}
- {%- else -%}
- {{entry.abbr}}
- {%- endif -%}
- {%- endif -%}
-