AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains

Anisha Gunjal, Anthony Wang, Elaine Lau, Vaskar Nath, Yunzhong He, Bing Liu, Sean M. Hendryx

ICLR 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 311 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
replace direct Likert reward with rubric-based implicit rewardGRPO on Qwen2.5-7B · HealthBenchoverall score: up to 31% relative gainImproved
replace direct Likert reward with rubric-based implicit rewardGRPO on Qwen2.5-7B · GPQA-Diamondaccuracy: up to 7% relative gainImproved
replace reference-Likert reward with rubric rewardGRPO on Qwen2.5-7B · HealthBench-1koverall score: 31.7% -> 35.9%Improved
use generic fixed rubrics instead of instance-specific rubrics (RaR-Predefined) (ablation)RaR · HealthBench / GPQA-DiamondscoreHurt
generate rubrics with access to reference answers (ablation)RaR-Implicit-Synthetic · HealthBench-1koverall score: 32.0% -> 35.9%Improved
add reference answers vs human-authored rubrics (ablation)RaR-Implicit · HealthBench-1koverall score: 34.8% (human) vs 35.9% (synthetic with ref)No clear effect
use all rubric criteria instead of Essential-only (ablation)RaR · HealthBench-1koverall score: 34.9% -> 37.2%Improved
add pitfall criteria (ablation)RaR · HealthBench-1koverall score: 37.2% -> 37.2%No clear effect
add categorical weight labels (ablation)RaR · HealthBench-1koverall score: 38.8% (none) vs 37.2% (labels)No clear effect
use a stronger LLM to generate reference-free rubrics (ablation)RaR · HealthBench-1koverall score: GPT-4o 34.2%, Qwen-32B 31.1%, Qwen-7B 31.9%Mixed
give judge rubrics (RaR-Implicit) instead of direct LikertQwen-7B-Instruct judge · synthetic medical judge qualityjudge accuracy: 22.0% -> 26.7%Improved
give judge rubrics instead of direct LikertQwen-14B-Instruct judge · synthetic medical judge qualityjudge accuracy: 24.9% -> 25.0%No 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