Hao Peng

Executable Counterfactuals: Improving LLMs' Causal Reasoning Through Code

mCLM: A Modular Chemical Language Model that Generates Functional and Makeable Molecules

From f(x) and g(x) to f(g(x)): LLMs Learn New Skills in RL by Composing Old Ones

Generalization of RLVR Using Causal Reasoning as a Testbed

Process Reinforcement through Implicit Rewards

Process Reward Models That Think

The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models

The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning

The Best Instruction-Tuning Data are Those That Fit

Retrieval Head Mechanistically Explains Long-Context Factuality

Reinforcement Learning Finetunes Small Subnetworks in Large Language Models

OpenHands: An Open Platform for AI Software Developers as Generalist Agents

Eliminating Position Bias of Language Models: A Mechanistic Approach

Advancing LLM Reasoning Generalists with Preference Trees

A Little Goes a Long Way: Efficient Long Context Training and Inference with Partial Contexts

Scaling Diffusion Language Models via Adaptation from Autoregressive Models

$PLUM$: Improving Code LMs Using On-Policy Preference Learning Powered by Automatic Test Cases

Free Process Rewards without Process Labels

Improving Influence-based Instruction Tuning Data Selection for Balanced Learning of Diverse Capabilities

S2-Attention: Hardware-Aware Context Sharding Among Attention Heads

FactCheckmate: Preemptively Detecting and Mitigating Hallucinations in LMs