11 papers
OS-Bench: Evaluating SWE Agents on Cross-OS Code Migration
Qirui Jin, Kenan Li, Liao Zhu +8
Language-model (LM) agents are increasingly evaluated on repository-level software engineering (SWE) benchmarks derived from GitHub issues and pull requests. Although these benchma…
Definition Grounding in Theory-Level Auto-formalization
Leyan Pan, Christian Szegedy, Frederick Rajasekaran +3
LLM-based auto-formalization systems can now generate formally verified Lean code at unprecedented rates. However, a formalization advances mathematical research only when further…
ChainSWE: Benchmarking Coding Agents on Multi-Bug Software Maintenance
Qirui Jin, Lingching Tung, Kenan Li +12
Language model (LM) agents are increasingly deployed to maintain codebases over extended periods, fixing streams of related defects while carrying context from one fix to the next.…
PACT: Privileged Trace Co-Training for Multi-Turn Tool-Use Agents
Zhenbang Du, Jun Luo, Zhiwei Zheng +8
Multi-turn tool-use agents must reason, call tools, and adapt to observations across several interaction turns. Post-training such agents is challenging, as reinforcement learning…
ORACLE-SWE: Quantifying the Contribution of Critical Information Signals on SWE Agents
Kenan Li, Qirui Jin, Liao Zhu +13
Recent advances in Language Model (LM) agents have increasingly focused on repository-level software engineering (SWE) tasks, leading to various agentic workflows, training methods…
CopT: Contrastive On-Policy Thinking with Continuous Spaces for General and Agentic Reasoning
Dachuan Shi, Hanlin Zhu, Xiangchi Yuan +4
Chain-of-thought (CoT) is a standard approach for eliciting reasoning capabilities from large language models (LLMs). However, the common CoT paradigm treats thinking as a prerequi…
MetaState: Persistent Working Memory Enhances Reasoning in Discrete Diffusion Language Models
Kejing Xia, Mingzhe Li, Lixuan Wei +5
Discrete diffusion language models (dLLMs) generate text by iteratively denoising a masked sequence. However, standard dLLMs condition each denoising step solely on the current har…
Behavior Knowledge Merge in Reinforced Agentic Models
Xiangchi Yuan, Dachuan Shi, Chunhui Zhang +4
Reinforcement learning (RL) is central to post-training, particularly for agentic models that require specialized reasoning behaviors. In this setting, model merging offers a pract…
Mitigating Forgetting Between Supervised and Reinforcement Learning Yields Stronger Reasoners
Xiangchi Yuan, Xiang Chen, Tong Yu +4
Large Language Models (LLMs) show strong reasoning abilities, often amplified by Chain-of-Thought (CoT) prompting and reinforcement learning (RL). Although RL algorithms can substa…
SwiReasoning: Switch-Thinking in Latent and Explicit for Pareto-Superior Reasoning LLMs
Dachuan Shi, Abedelkadir Asi, Keying Li +4
Recent work shows that, beyond discrete reasoning through explicit chain-of-thought steps, which are limited by the boundaries of natural languages, large language models (LLMs) ca…
Can Transformers Reason Logically? A Study in SAT Solving
Leyan Pan, Vijay Ganesh, Jacob Abernethy +2
We formally study the logical reasoning capabilities of decoder-only Transformers in the context of the boolean satisfiability (SAT) problem. First, we prove by construction that d…