第17回最先端NLP勉強会

8月31日(日) プログラム

[09:30-10:00]  受付
[10:00-10:10]  オープニング

• 清野舜: No Need to Talk: Asynchronous Mixture of Language Models (ICLR2025; Anastasiia Filippova, Angelos Katharopoulos, David Grangier, Ronan Collobert) [paper] [slide]
• 栗田宙人: Learning to (Learn at Test Time): RNNs with Expressive Hidden States (ICML2025; Yu Sun, Xinhao Li, Karan Dalal, Jiarui Xu, Arjun Vikram, Genghan Zhang, Yann Dubois, Xinlei Chen, Xiaolong Wang, Sanmi Koyejo, Tatsunori Hashimoto, Carlos Guestrin) [paper] [slide]
• 大葉大輔: Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models (ICLR2025; Marianne Arriola, Aaron Gokaslan, Justin T Chiu, Zhihan Yang, Zhixuan Qi, Jiaqi Han, Subham Sekhar Sahoo, Volodymyr Kuleshov) [paper] [slide]

• 西田光甫: Not All Tokens Are What You Need for Pretraining (NeurIPS2024; Zhenghao Lin, Zhibin Gou, Yeyun Gong, Xiao Liu, yelong shen, Ruochen Xu, Chen Lin, Yujiu Yang, Jian -Jiao, Nan Duan, Weizhu Chen) [paper] [slide]
• 高瀬翔: Spike No More: Stabilizing the Pre-training of Large Language Models (COLM 2025; Sho Takase, Shun Kiyono, Sosuke Kobayashi, Jun Suzuki) [paper] [slide]
• 中石海: Physics of Language Models: Part 3.3, Knowledge Capacity Scaling Laws (ICLR2025; Zeyuan Allen-Zhu, Yuanzhi Li) [paper] [slide]

• 矢野千紘: How Do Large Language Models Acquire Factual Knowledge During Pretraining? (NeurIPS 2024 (Poster); Hoyeon Chang, Jinho Park, Seonghyeon Ye, Sohee Yang, Youngkyung Seo, Du-Seong Chang, Minjoon Seo) [paper] [slide]
• 山本悠士: AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR2025; Junfeng Fang, Houcheng Jiang, Kun Wang, Yunshan Ma, Jie Shi, Xiang Wang, Xiangnan He, Tat-Seng Chua) [paper] [slide]
• 井之上直也: Do I Know This Entity? Knowledge Awareness and Hallucinations in Language Models (ICLR2025; Javier Ferrando, Oscar Balcells Obeso, Senthooran Rajamanoharan, Neel Nanda) [paper] [slide]

• 王昊: Language Model Alignment in Multilingual Trolley Problems (ICLR2025; Zhijing Jin, Max Kleiman-Weiner, Giorgio Piatti, Sydney Levine, Jiarui Liu, Fernando Gonzalez Adauto, Francesco Ortu, András Strausz, Mrinmaya Sachan, Rada Mihalcea, Yejin Choi, Bernhard Schölkopf) [paper] [slide]
• 水木栄: Learning Dynamics of LLM Finetuning (ICLR2025; Yi Ren, Danica J. Sutherland) [paper] [slide]
• 李凌寒: Intuitive Fine-Tuning: Towards Simplifying Alignment into a Single Process (ACL2025; Ermo Hua, Biqing Qi, Kaiyan Zhang, Kai Tian, Xingtai Lv, Ning Ding, Bowen Zhou) [paper] [slide]

• 銭本友樹: Can Language Models Reason about Individualistic Human Values and Preferences? (ACL2025; Liwei Jiang, Taylor Sorensen, Sydney Levine, Yejin Choi) [paper] [slide]
• 岡崎直観: BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions (ICLR2025; Terry Yue Zhuo, Vu Minh Chien, Jenny Chim, Han Hu, Wenhao Yu, Ratnadira Widyasari, Imam Nur Bani Yusuf, Haolan Zhan, Junda He, Indraneil Paul, Simon Brunner, Chen GONG, and many more) [paper] [slide]
• 木山朔: People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text (ACL2025; Jenna Russell, Marzena Karpinska, Mohit Iyyer) [paper] [slide]
• 亀田尭宙: INCLUDE: Evaluating Multilingual Language Understanding with Regional Knowledge (ICLR2025; Angelika Romanou, Negar Foroutan, Anna Sotnikova, Sree Harsha Nelaturu, Shivalika Singh, Rishabh Maheshwary, Micol Altomare, Zeming Chen, Mohamed A. Haggag, and many more) [paper] [slide]

9月1日(月) プログラム

[09:30-10:00]  受付
[10:00-10:10]  オープニング

• 持橋大地: Training Large Language Models to Reason in a Continuous Latent Space (COLM 2025; Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston, Yuandong Tian) [paper] [slide]
• 勝見久央: ReGenesis: LLMs can Grow into Reasoning Generalists via Self-Improvement (ICLR2025; XIANGYU PENG, Congying Xia, Xinyi Yang, Caiming Xiong, Chien-Sheng Wu, Chen Xing) [paper] [slide]
• 井田龍希: Disentangling Memory and Reasoning Ability in Large Language Models (ACL2025; Mingyu Jin, Weidi Luo, Sitao Cheng, Xinyi Wang, Wenyue Hua, Ruixiang Tang, William Yang Wang, Yongfeng Zhang) [paper] [slide]

• 篠田一聡: Safety Alignment Should be Made More Than Just a Few Tokens Deep (ICLR2025; Xiangyu Qi, Ashwinee Panda, Kaifeng Lyu, Xiao Ma, Subhrajit Roy, Ahmad Beirami, -Prateek Mittal, Peter Henderson) [paper] [slide]
• 大井聖也: Training Language Models to Self-Correct via Reinforcement Learning (ICLR2025; Aviral Kumar, Vincent Zhuang, Rishabh Agarwal, Yi Su, John D Co-Reyes, Avi Singh, Kate Baumli, Shariq Iqbal, Colton Bishop, Rebecca Roelofs, Lei M Zhang, Kay McKinney, Disha Shrivastava, Cosmin Paduraru, George Tucker, Doina Precup, Feryal Behbahani, Aleksandra Faust) [paper] [slide]
• 塩野大輝: Words or vision: Do vision-language models have blind faith in text? (CVPR2025; Ailin Deng, Tri Cao, Zhirui Chen, Bryan Hooi) [paper] [slide]

• 笹野遼平: The Impact of Token Granularity on the Predictive Power of Language Model Surprisal (ACL2025; Byung-Doh Oh, William Schuler) [paper] [slide]
• 西田悠人: What Goes Into a LM Acceptability Judgment? Rethinking the Impact of Frequency and Length (NAACL2025; Lindia Tjuatja, Graham Neubig, Tal Linzen, Sophie Hao) [paper] [slide]
• 辻村有輝: Quantifying Semantic Emergence in Language Models (ACL2025; Hang Chen, Xinyu Yang, Jiaying Zhu, Wenya Wang) [paper] [slide]

• 根石将人: Cramming 1568 Tokens into a Single Vector and Back Again: Exploring the Limits of Embedding Space Capacity (ACL2025; Yuri Kuratov, Mikhail Arkhipov, Aydar Bulatov, Mikhail Burtsev) [paper] [slide]
• 横井祥: Language Models Are Implicitly Continuous (ICLR 2025 (Poster); Samuele Marro, Davide Evangelista, X. Angelo Huang, Emanuele La Malfa, Michele Lombardi, Michael J. Wooldridge) [paper] [slide]
• 森友亮: From Language Models over Tokens to Language Models over Characters (ICML2025; Tim Vieira, Benjamin LeBrun, Mario Giulianelli, Juan Luis Gastaldi, Brian DuSell, John Terilla, Timothy J. O'Donnell, Ryan Cotterell) [paper] [slide]

• 神藤駿介: Constructing language: a framework for explaining acquisition (Trends in Cognitive Sciences; Caroline F. Rowland, Gert Westermann, Anna L. Theakston, Julian M. Pine, Padraic Monaghan, and Elena V.M. Lieven) [paper] [slide]
• 森下睦: Investigating Critical Period Effects in Language Acquisition through Neural Language Models (TACL2025; Ionut Constantinescu, Tiago Pimentel, Ryan Cotterell, Alex Warstadt) [paper] [slide]
• 吉田遼 [slide]
  • The Harmonic Structure of Information Contours (ACL2025; Eleftheria Tsipidi, Samuel Kiegeland, Franz Nowak, Tianyang Xu, Ethan Wilcox, Alex Warstadt, Ryan Cotterell, Mario Giulianelli) [paper] 
  • Between Circuits and Chomsky: Pre-pretraining on Formal Languages Imparts Linguistic Biases (ACL2025; Michael Y. Hu, Jackson Petty, Chuan Shi, William Merrill, Tal Linzen) [paper]