00

GRPO with State Mutation

Technique ☆ WATCH ⎘ SHARE

First seen 9/3/2026

Last seen 9/3/2026

Evidence 8 chunks

01

NEIGHBORHOOD

8 nodes · 11 edges

graph · GRPO with State Mutation · depth=1

02

RELATIONSHIPS

7 connections

advantage computation uses → 100% 2e evidence

GRPO-SMu modifies the advantage computation to handle sparse rewards.

Group Relative Policy Optimization extends → 100% 2e evidence

GRPO-SMu modifies and improves upon conventional GRPO by diversifying input states.

tree-based branching mutation strategy uses → 100% 2e evidence

GRPO-SMu relies on the tree-based branching mutation strategy to construct its training data.

GRPO with State Mutations: Improving LLM-Based Hardware Test Plan Generation ← introduces 100% 2e evidence

The paper proposes GRPO-SMu as a novel RL training method.

reinforcement learning for language models implements → 100% 2e evidence

GRPO-SMu is a specific implementation of reinforcement learning for language model training.

DeepSeek-R1-distill-Qwen-7B ← uses 100% 2e evidence

DeepSeek-R1-distill-Qwen-7B is trained with GRPO-SMu to achieve best results.

sequential circuit verification uses → 90% 1e evidence

GRPO-SMu shows particular improvement for sequential circuit verification.