7 papers
Continual Memorization of Factoids in Language Models
Howard Chen, Jiayi Geng, Adithya Bhaskar +2
As new knowledge rapidly accumulates, language models (LMs) with pretrained knowledge quickly become obsolete. A common approach to updating LMs is fine-tuning them directly on new…
Self-Distillation Zero: Self-Revision Turns Binary Rewards into Dense Supervision
Yinghui He, Simran Kaur, Adithya Bhaskar +7
Current post-training methods in verifiable settings fall into two categories. Reinforcement learning (RLVR) relies on binary rewards, which are broadly applicable and powerful, bu…
Extracting Rule-based Descriptions of Attention Features in Transformers
Dan Friedman, Adithya Bhaskar, Alexander Wettig +1
Mechanistic interpretability strives to explain model behavior in terms of bottom-up primitives. The leading paradigm is to express hidden states as a sparse linear combination of…
Language Models that Think, Chat Better
Adithya Bhaskar, Xi Ye, Danqi Chen
Reinforcement learning with verifiable rewards (RLVR) trains language models to use long chain-of-thought reasoning (CoT) in domains like mathematics and code with rule-based verif…
Cache Me If You Can: How Many KVs Do You Need for Effective Long-Context LMs?
Adithya Bhaskar, Alexander Wettig, Tianyu Gao +2
Language models handle increasingly long contexts for tasks such as book summarization, but this leads to growing memory costs for the key-value (KV) cache. Many prior works have p…
Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization
Noam Razin, Sadhika Malladi, Adithya Bhaskar +3
Direct Preference Optimization (DPO) and its variants are increasingly used for aligning language models with human preferences. Although these methods are designed to teach a mode…