Linyi Yang

7

Papers

159

Total Citations

Papers (7)

CycleResearcher: Improving Automated Research via Automated Review

ReMA: Learning to Meta-Think for LLMs with Multi-agent Reinforcement Learning

NeurIPS 2025arXiv

MMQA: Evaluating LLMs with Multi-Table Multi-Hop Complex Questions

CofCA: A STEP-WISE Counterfactual Multi-hop QA benchmark

ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning

NeurIPS 2025arXiv

Constrain Alignment with Sparse Autoencoders

USB: A Unified Semi-supervised Learning Benchmark for Classification