AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Draft-and-Audit Reinforcement Learning for Optimization Modeling

Zeping Min, Weihang Xu, Zhengzhong You, Wotao Yin, Xinshang Wang

ICML 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards

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 the Draft-Audit prompt to the base model without RL (ablation)Qwen2.5-7B-Instruct · MAMO-Complex / IndustryORpass@1 accuracy (%): +2.4/+0.0 ppNo clear effect
add an audit prompt to single-turn RL model (ablation)single-turn RL · MAMO-Complex / IndustryORpass@1 accuracy (%): 76.7 -> 79.1 and 20.0 -> 20.0No clear effect
train two-turn Draft-Audit policy instead of single-turn RL (ablation)DA-RL · MAMO-Complex / IndustryORpass@1 accuracy (%): +7.8/+16.7 ppImproved
train with the audit turn so Turn-1 drafts improve (ablation)DA-RL Turn-1 draft · MAMO-Complex / IndustryORpass@1 accuracy (%): +3.9/+15.0 pp over single-turn RLImproved
double generation length for single-turn RL (ablation)single-turn RL · MAMO-Complex / IndustryORpass@1 accuracy (%): 76.7/20.0No clear effect
use structured audit rubric instead of generic prompt (ablation)DA-RL · MAMO-Complexpass@1 accuracy (%): 79.1 -> 84.5 (-5.4 without)Improved
use structured audit rubric instead of generic prompt (ablation)DA-RL · IndustryORpass@1 accuracy (%): 20.0 -> 36.7 (-16.7 without)Improved
extend rollout from two to three turns (ablation)DA-RL · MAMO-Complex / IndustryORpass@1 accuracy (%): 84.5 -> 74.4 and 36.7 -> 15.0 (-10.1/-21.7)Hurt
add few-shot examples to the prompts (ablation)DA-RL · MAMO-Complexpass@1 accuracy (%): 76.7% -> 84.5%Improved
add few-shot examples to the prompts (ablation)DA-RL · OptMATH-Benchpass@1 accuracy (%): 11.8% -> 30.4%Improved
add Draft-Audit prompting to frontier LLMsGemini-3-Pro · Macro AVGpass@1 accuracy (%): 8.4 -> 42.1 (+33.7)Improved
add Draft-Audit prompting to frontier LLMsClaude-Opus-4.5 · Macro AVGpass@1 accuracy (%): 57.4 -> 67.0 (+9.6)Improved

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