AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

When Rubrics Fail: Error Enumeration as Reward for Reference-Free RL Post-Training

Wisdom Ikezogwo, Mehmet Saygin Seyfioglu, Ranjay Krishna, Karim Bouyarmane

ECCV 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 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
use implicit error counting reward instead of rubric reward (RaR)GRPO-trained Kontext VTO · MDressBench (flat references)CEC: 5.60 -> 5.31Improved
use implicit error counting reward instead of rubric reward (RaR)GRPO-trained Kontext VTO · MDressBench (non-flat references)CEC: 5.533 -> 5.203Improved
use implicit error counting instead of direct scalar scoringGRPO-trained Kontext VTO · MDressBench (flat references)CEC: 5.373 -> 5.312Improved
add RL post-training with RaR rubric reward vs direct scoringGRPO-trained Kontext VTO · MDressBench (flat references)CEC: 5.373 -> 5.596Hurt
use explicit error counting (EEC) instead of implicit (ablation)IEC reward · MDressBench (non-flat)CEC: 5.20 -> 5.72Hurt
use explicit error counting (EEC) instead of implicit (ablation)IEC reward · MDressBench (flat references)CEC: 5.31 -> 6.03Hurt
add group calibration (ablation)IEC reward · MDressBench (non-flat references)CEC: 5.259 -> 5.203Improved
add group calibration (ablation)IEC reward · VITON-HDFID: 5.190 -> 5.120Improved
add group calibration (ablation)IEC reward · DressCodeFID: 3.980 -> 3.610Improved
use CEC as selector instead of Direct or RaRHuman preference selection · Human study (25 samples)Top-1 accuracy: 0.30 -> 0.60Improved
increase best-of-n from 8 to 12 (ablation)IEC GRPO · MDressBench (flat)CEC: 5.31 -> 4.80Improved
increase best-of-n from 8 to 12 (ablation)IEC GRPO · MDressBench (non-flat)CEC: 5.20 -> 5.27Hurt

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