AI Paper Atlas

Research map / Large language models / Alignment and preferences

Learning Ordinal Probabilistic Reward from Preferences

Longze Chen, Lu Wang, Renke Shan, Ze Gong, Run Luo, Jiaming Li, Jing Luo, Qiyao Wang, Min Yang

ICLR 2026 · Large language models › Alignment and preferences · approach: Reward model design and preference models · 2 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 the OPRM ordinal loss instead of Bradley-Terry loss (ablation)Qwen-32B reward model · RM benchmarksaccuracy: consistent performance penalty of 1.1% to 2.1% when replaced by BTImproved
use weighted-average decoding instead of Argmax (ablation)OPRM · RM benchmarksaccuracy: Argmax leads to 8.7% to 17.5% dropImproved
add region flooding tuning (RgFT)OPRM-32B · RM benchmarks OverallOverall: 74.2 -> 73.9Hurt
add region flooding tuning (RgFT)OPRM-32B · PPE-Correctnessaccuracy: 66.1 -> 67.3Improved
add region flooding tuning (RgFT)OPRM-72B · RM benchmarks OverallOverall: 73.1 -> 73.8Improved
add region flooding tuning (RgFT)OPRM-32B · RewardBenchECE-10: 10.62 -> 5.18Improved
use OPRM instead of base Qwen2.5-32BQwen-2.5-32B · RewardBenchaccuracy: 69.56 -> 81.04Improved
add RgFT with unlabeled general-domain data (mixed training)OPRM-RgFT · Role PlayOverall: 65.4 -> 70.2Improved
add general-domain dataOPRM · Role PlayOverall: 64.3 -> 68.1Improved
use region flooding instead of hard classification (ablation)OPRM-RgFT-32B · RM benchmarksOverall: 54.3 -> 73.9Improved
use uniform instead of irregular bin boundaries (ablation)OPRM-RgFT · RM benchmarksOverall: 73.8 -> 73.9No clear effect
scale model from 32B to 72BOPRM · RM benchmarksOverall: 74.2 -> 73.1Hurt

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