Pengfei Liu

SR-Scientist: Scientific Equation Discovery With Agentic AI

MegaScience: Pushing the Frontiers of Open Post-Training Datasets for Science Reasoning

InnovatorBench: Evaluating Agents’ Ability to Conduct Innovative AI Research

ResearcherBench: Evaluating Deep AI Research Systems on the Frontiers of AI Research

GeneVLM: Automated Parsing Executable Digital Gene from a Single Image

LIMI: Less is More for Agency

Efficient Agent Training for Computer Use

Proximal Supervised Fine-Tuning

DatasetResearch: Benchmarking Agent Systems for Demand-Driven Dataset Discovery

One RL to See Them All: Visual Triple Unified Reinforcement Learning

Discovering Architectures via an Evolutionary Agentic Framework

Visual Programmability: A Guide for Code-as-Thought in Chart Understanding

Attention Localization Through Separator Tokens: Unlocking Long Numerical Sequence Processing in LLMs

ARGO: Asynchronous Rollout with Human Guidance for Research Agent Optimization

Deep Cognition: A Multi-Agent Framework for Collaborative Research with Real-Time Cognitive Oversight

One Sample to Rule Them All: Extreme Data Efficiency in RL Scaling

Weak-to-Strong Preference Optimization: Stealing Reward from Weak Aligned Model

Programming Every Example: Lifting Pre-training Data Quality Like Experts at Scale

Progress or Regress? Self-Improvement Reversal in Post-training

On Evaluating LLM Alignment by Evaluating LLMs as Judges

LIMOPro: Reasoning Refinement for Efficient and Effective Test-time Scaling

RepoST: Scalable Repository-Level Coding Environment Construction with Sandbox Testing

LIMO: Less is More for Reasoning

Programming Every Example: Lifting Pre-training Data Quality like Experts at Scale

OmniBal: Towards Fast Instruction-Tuning for Vision-Language Models via Omniverse Computation Balance

FacTool: Factuality Detection in Generative AI -- A Tool Augmented Framework for Multi-Task and Multi-Domain Scenarios

BeHonest: Benchmarking Honesty in Large Language Models

OMNIBAL: TOWARDS FAST INSTRUCT-TUNING FOR VISION-LANGUAGE MODELS VIA OMNIVERSE COMPUTATION BALANCE

CodeBenchGen: Creating Scalable Execution-based Code Generation Benchmarks

TOMVALLEY: EVALUATING THE THEORY OF MIND REASONING OF LLMS IN REALISTIC SOCIAL CONTEXT