mlsanityData to 7 Oct 2026

Execution-Grounded Credit Assignment for GRPO in Code Generation

arXiv 2603.16158 ↗First posted 17 Mar 2026cs.LGcited by 5

On the map: LLM reasoning → Reinforcement learning for reasoning

Abstract (arXiv)

Critic-free reinforcement learning with verifiable rewards (RLVR) improves code generation by optimizing unit-test pass rates, but GRPO-style updates suffer from coarse credit assignment: a single outcome signal is spread uniformly across long programs even when failure stems from a localized semantic error. We propose Execution-Grounded Credit Assignment (EGCA), which localizes GRPO updates using execution traces. For programs that satisfy algorithmic constraints but fail tests, EGCA executes the candidate and a canonical reference solution (curated once offline; used for analysis, not supervision) under identical instrumentation, identifies the earliest semantic divergence, and assigns advantage only to the corresponding token span while masking downstream tokens. EGCA is a drop-in modification requiring no critic, auxiliary loss, or learned verifier, yielding 82.1% pass@1 on HumanEval (+3.1 over GRPO) and 68.9% on MBPP (+1.5) with 18% wall-clock overhead.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning. Builds on 9 of its 16 references that are in the map. First posted on arXiv 17 Mar 2026.

Builds on

9 of 16 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025Exploiting Tree Structure for Credit Assignment in RL Training of LLMs22 Sep
  2. 2023Let's Verify Step by Step31 May

Other references · 7

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2021Program Synthesis with Large Language Modelscited by 3,509
  4. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,154
  5. 2024Qwen2.5-Coder Technical Reportcited by 1,151
  6. 2023Execution-based Code Generation using Deep Reinforcement Learningcited by 95
  7. 2025CodeRL+: Improving Code Generation via Reinforcement with Execution Semantics Alignmentcited by 22

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Unifying Group-Relative and Self-Distillation Policy Optimization via Sample Routingcited by 65
  2. 2026Knowing When to Ask: Segment-Level Credit Assignment for LLM Tool Usecited by 2
  3. 2026DRIFT: Difficulty Routing Self-DIstillation with Rhythm-Gated Exploration and Success BuFfer Trainingcited by 2
  4. 2026MoRSE: Task-Oriented Multi-Agent System with Mixture of Role-Subtask Expertscited by 2
  5. 2026When Implausible Tokens Get Reinforced: Tail-Aware Credit Calibration for LLM Reinforcement Learningcited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2025MURPHY: Feedback-Aware GRPO with Retrospective Credit Assignment for Multi-Turn Code Generation
  2. 2026ExecuCritic: Calibrated Critic Shaping for Code Generation with Verifiable Rewards ↗
  3. 2026StepCodeReasoner: Aligning Code Reasoning with Stepwise Execution Traces via Reinforcement Learning
  4. 2026Rubric-to-Code Credit Assignment for Reinforcement Learning
  5. 2026BoT-GRPO: Efficient Process-Reward RL for Reasoning via Bag-of-Token Aggregation
  6. 2025From Reasoning to Code: GRPO Optimization for Underrepresented Languages
  7. 2026Reinforcement Learning for Code Optimization
  8. 2025Afterburner: Reinforcement Learning Facilitates Self-Improving Code Efficiency Optimization