Publications

ICML 2026

RLAnything: Forge Environment, Policy, and Reward Model in Completely Dynamic RL System

General agentic RL · corresponding-author work

Paper Code

ICML 2026Spotlight · Top 3%

Latent Collaboration in Multi-Agent Systems (LatentMAS)

Efficient multi-agent reasoning in latent space · corresponding-author work

Paper Code

TECH REPORT

OpenClaw-RL: Train Any Agent Simply by Talking

Fully asynchronous RL for interactive agents

Paper Code

ICLR 2026

Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models (TraceRL)

Trajectory-aware RL for diffusion LLMs · corresponding-author work

Paper Code

NEURIPS 2025

MMaDA: Multimodal Large Diffusion Language Models

Unified multimodal generative modeling

Paper Code

NEURIPS 2025Spotlight · Top 3%

Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning (ReasonFlux-Coder)

Coder / unit-tester co-evolution via RL · corresponding-author work

Paper Code

NEURIPS 2025

ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs

Trajectory-aware process reward models · corresponding-author work

Paper Code

PREPRINT

ReasonFlux: Hierarchical LLM Reasoning via Scaling Thought Templates

Structured reasoning and post-training

Paper Code

NEURIPS 2024Spotlight · Top 3%

Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models

Structured reasoning with thought templates

Paper Code

ICML 2024

Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs

Generative modeling foundation (RPG)

Paper Code

Work across models, agents, RL, and discovery.

RLAnything: Forge Environment, Policy, and Reward Model in Completely Dynamic RL System

Latent Collaboration in Multi-Agent Systems (LatentMAS)

OpenClaw-RL: Train Any Agent Simply by Talking

Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models (TraceRL)

MMaDA: Multimodal Large Diffusion Language Models

Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning (ReasonFlux-Coder)

ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs

ReasonFlux: Hierarchical LLM Reasoning via Scaling Thought Templates

Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models

Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs

70+ papers across machine learning and AI.