AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learning

Xin Cheng, Shuo He, Lang Feng, Haiyang Xu, Ming Yan, Lei Feng, Bo An

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add episode-level advantage A^E to step-level advantage (ablation)GraphGPO · ALFWorldsuccess rate (%): 91.67 -> 92.71Improved
add episode-level advantage A^E to step-level advantage (ablation)GraphGPO · Sokobansuccess rate (%): 83.07 -> 86.98Improved
add episode-level advantage A^E to step-level advantage (ablation)GiGPO · Sokobansuccess rate (%): 62.5 -> 76.92Improved
add episode-level advantage A^E to step-level advantage (ablation)GiGPO · ALFWorldsuccess rate (%): 91.41 -> 90.88No clear effect
add dynamic samplingGraphGPO · ALFWorldsuccess rate (%): 92.71 -> 98.43Improved
add dynamic samplingGraphGPO · WebShopsuccess rate (%): 78.65 -> 85.68Improved
add dynamic samplingGiGPO · ALFWorldsuccess rate (%): 90.88 -> 96.35Improved
use graph-based advantage instead of anchor-state GiGPO advantageGiGPO · ALFWorld Qwen2.5-1.5Bsuccess rate (%): 90.88 -> 92.71Improved
use graph-based advantage instead of GRPO trajectory-level advantageGRPO · ALFWorld Qwen2.5-7Bsuccess rate (%): 83.33 -> 95.31Improved
use graph-based advantage instead of GiGPOGiGPO · Sokobansuccess rate (%): 76.92 -> 86.98Improved
vary distance discount omega from 0.1 to 0.6 (ablation)GraphGPO · ALFWorld (1.5B)success rate (%): 92.71 -> 90.36No clear effect
vary distance discount omega from 0.2 to 0.8 (ablation)GraphGPO · WebShop (1.5B)success rate (%): 78.65 -> 75.00No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy