AI Paper Atlas

Research map / Large language models / Alignment and preferences

The Hidden Link between RLHF and Contrastive Learning

Xufei Lv, Kehai Chen, Haoyuan Sun, Xuefeng Bai, Min Zhang, Houde Liu

ICML 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO · 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
replace DV/MINE bound (DPO) with JS estimator (MIO)Mistral-7B-SFT · Arena-Hardscore: 16.3 -> 23.6Improved
replace DPO with MIOMistral-7B-SFT · AlpacaEval 2win rate: 16.3 -> 21.2Improved
replace DPO with MIOMistral-7B-SFT · MT-Benchscore: 7.3 -> 7.3No clear effect
replace DPO with MIOQwen2.5-7B-Instruct · GSM8Kaccuracy: 73.46 -> 75.39Improved
replace DPO with MIOLLaMA3-8B-SFT · MUSRaccuracy: 39.11 -> 42.81Improved
replace DPO with MIO on rare-rejected examplesQwen2.5-7B-Instruct · Local update diagnostic, buckets 0-6change in chosen likelihood: DPO -0.1825 vs MIO 3.3761 (bucket 0)Improved
replace DPO with MIO in toy modeltoy MLP policy · toy model, rejected prob near zerochosen/rejected likelihood dynamicsImproved
use MIO instead of DPO in online preference optimizationQwen3.5-9B · TL;DR onlinewin rate: 30.1 -> 35.5Improved
replace DPO with MIOGPT-J SFT · CNN/DailyMail OOD (temp 0)GPT-4 win rate: 0.36 -> 0.38Improved
increase beta from 2.0 to 20.0 (ablation)MIO, Mistral-7B-SFT · IFEvalscore: 46.37 -> 36.59Hurt
vary beta (ablation)MIO, Mistral-7B-SFT · GSM8Kaccuracy: 39.91 (beta 2.0), 40.16 (beta 10.0), 38.74 (beta 20.0)No clear effect
add 10% label noise to training data (ablation)MIO vs DPO, Qwen2.5-7B-Instruct · MUSRaccuracy: DPO 38.5 -> 36.7, MIO 38.5 -> 36.0Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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