6 citations · 17 across the 24 of their papers we have counts for
26 papers
Agent Plasticity: Measuring Self-Improvement Through Experience
Harman Singh, Anton Bakhtin, Rulin Shao +8
AI agents increasingly operate in environments where they can diagnose failures and improve through experience, yet existing evaluations largely measure what an agent can do at a f…
GitSwarm: Decentralized Compounding Inference
Vedant Shah, Ankur Samanta, Paras Dahal +11
Long-horizon problem solving and scientific research require computation to accumulate across successive attempts. Partial solutions, experimental findings, and unsuccessful approa…
Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning
Yinghui He, Ling Yang, Jiarui Liu +6
Long-horizon reasoning in recent LLMs demands that the model switch between distinct skills inside a reasoning chain, such as first doing a math derivation, then using the result t…
T^2MLR: Transformer with Temporal Middle-Layer Recurrence
Ziyang Cai, Xingyu Zhu, Yihe Dong +2
Transformer reasoning is limited by autoregressive decoding, which repeat edly compresses rich hidden computation through token space and makes it difficult for intermediate reason…
Rethinking On-Policy Self-Distillation for Thinking Models
Simran Kaur, Narutatsu Ri, Yinghui He +2
Self-distillation is a promising recipe for self-improvement in language models. In this setting, a model can serve as its own teacher when given privileged information, such as a…
When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient
Shuning Shang, Hubert Strauss, Stanley Wei +2
Training language models via reinforcement learning often relies on imperfect proxy rewards, since ground truth rewards that precisely define the intended behavior are rarely avail…
Scaling Test-Time Compute for Agentic Coding
Joongwon Kim, Wannan Yang, Kelvin Niu +13
Test-time scaling has become a powerful way to improve large language models. However, existing methods are best suited to short, bounded outputs that can be directly compared, ran…
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…
Contextual Drag: How Errors in the Context Affect LLM Reasoning
Yun Cheng, Xingyu Zhu, Haoyu Zhao +1
Central to many self-improvement pipelines for large language models (LLMs) is the assumption that models can improve by reflecting on past mistakes. We study a phenomenon termed c…
Escaping the Cognitive Well: Efficient Competition Math with Off-the-Shelf Models
Xingyu Dang, Rohit Agarwal, Rodrigo Porto +3
In the past year, custom and unreleased math reasoning models reached gold medal performance on the International Mathematical Olympiad (IMO). Similar performance was then reported…
How Does RL Post-training Induce Skill Composition? A Case Study on Countdown
Simon Park, Simran Kaur, Sanjeev Arora
While reinforcement learning (RL) successfully enhances reasoning in large language models, its role in fostering compositional generalization (the ability to synthesize novel skil…
On the Impossibility of Retrain Equivalence in Machine Unlearning
Jiatong Yu, Yinghui He, Anirudh Goyal +1
Machine unlearning seeks to selectively remove the "influence" of specific training data on a model's outputs. The ideal goal is Retrain Equivalence--behavior identical to a model…
Rethinking Thinking Tokens: LLMs as Improvement Operators
Lovish Madaan, Aniket Didolkar, Suchin Gururangan +6
Reasoning training incentivizes LLMs to produce long chains of thought (long CoT), which among other things, allows them to explore solution strategies with self-checking. This res…
Metacognitive Reuse: Turning Recurring LLM Reasoning Into Concise Behaviors
Aniket Didolkar, Nicolas Ballas, Sanjeev Arora +1
Large language models (LLMs) now solve multi-step problems by emitting extended chains of thought. During the process, they often re-derive the same intermediate steps across probl…
Weak-to-Strong Generalization Even in Random Feature Networks, Provably
Marko Medvedev, Kaifeng Lyu, Dingli Yu +3
Weak-to-Strong Generalization (Burns et al., 2024) is the phenomenon whereby a strong student, say GPT-4, learns a task from a weak teacher, say GPT-2, and ends up significantly ou…
On the Power of Context-Enhanced Learning in LLMs
Xingyu Zhu, Abhishek Panigrahi, Sanjeev Arora
We formalize a new concept for LLMs, context-enhanced learning. It involves standard gradient-based learning on text except that the context is enhanced with additional data on whi…
Unrealized Expectations: Comparing AI Methods vs Classical Algorithms for Maximum Independent Set
Yikai Wu, Haoyu Zhao, Sanjeev Arora
AI methods, such as generative models and reinforcement learning, have recently been applied to combinatorial optimization (CO) problems, especially NP-hard ones. This paper compar…
Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?
Simon Park, Abhishek Panigrahi, Yun Cheng +3
Vision Language Models (VLMs) are impressive at visual question answering and image captioning. But they underperform on multi-step visual reasoning -- even compared to LLMs on the…
Can Models Learn Skill Composition from Examples?
Haoyu Zhao, Simran Kaur, Dingli Yu +2
As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not enco…
Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning
Simran Kaur, Simon Park, Anirudh Goyal +1
We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data for instruction-following. The pipeline involves two stages, each leveraging an ex…
AI-Assisted Generation of Difficult Math Questions
Vedant Shah, Dingli Yu, Kaifeng Lyu +8
Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math que…
Metacognitive Capabilities of LLMs: An Exploration in Mathematical Problem Solving
Aniket Didolkar, Anirudh Goyal, Nan Rosemary Ke +7
Metacognitive knowledge refers to humans' intuitive knowledge of their own thinking and reasoning processes. Today's best LLMs clearly possess some reasoning processes. The paper g…
Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates
Kaifeng Lyu, Haoyu Zhao, Xinran Gu +3
Public LLMs such as the Llama 2-Chat underwent alignment training and were considered safe. Recently Qi et al. [2024] reported that even benign fine-tuning on seemingly safe datase…
Unlearning via Sparse Representations
Vedant Shah, Frederik Träuble, Ashish Malik +5
Machine \emph{unlearning}, which involves erasing knowledge about a \emph{forget set} from a trained model, can prove to be costly and infeasible by existing techniques. We propose…