ForkGRPO: Efficient Exploration and Fine-Grained Credit Assignment for Diffusion Language Models

arXiv:2610.19408 · cs.CL · Submitted 2026-10-14 · Read on arXiv

cs.CL

Submitted: 2026-10-14

Updated: 2026-10-14

Comments: 20 pages, 5 figures. The first two authors contributed equally

Code: https://github.com/LeapLabTHU/JustGRPO

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Credit assignment — 1× in this paper · explained in 3 episodes

Sources

• Program Synthesis with Large Language Models
• TreeAdv: Tree-Structured Advantage Redistribution for Group-Based RL
• Evaluating Large Language Models Trained on Code
• Training Verifiers to Solve Math Word Problems
• Reinforcement Learning for Diffusion LLMs with Entropy-Guided Step Selection and Stepwise Advantages
• Diffusion-State Policy Optimization for Masked Diffusion Language Models
• Back on Track: Aligning Rewards and States for Reasoning in Diffusion Large Language Models
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• Dream 7B: Diffusion Large Language Models
• Stabilizing Reinforcement Learning for Diffusion Language Models
• Enhancing Reasoning for Diffusion LLMs via Distribution Matching Policy Optimization

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering