AI Paper Atlas

Research map / Large language models / Alignment and preferences

Semi-Supervised Preference Optimization with Limited Feedback

Seonggyun Lee, Sungjun Lim, Seojin Park, Soeun Cheon, Kyungwoo Song

ICLR 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO · 1 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
add pseudo-labeled unpaired UltraChat data (SSPO) vs best baseline, 1% pairedMistral-7B · AlpacaEval 2.0 LC (UltraFeedback) 1%LC win rate: 18.2 -> 26.7Improved
SSPO with 1% paired vs baselines with 10% pairedMistral-7B · AlpacaEval 2.0 LC (UltraFeedback)LC win rate: 19.1 -> 26.7Improved
SSPO vs best baseline at 10% pairedLlama3-8B · AlpacaEval 2.0 LC (UltraFeedback) 10%LC win rate: 16.7 -> 20.7Improved
SSPO vs best baseline at 1% pairedLlama3-8B · AlpacaEval 2.0 LC (UltraFeedback) 1%LC win rate: 14.9 -> 14.8No clear effect
SSPO on MT-Bench vs baselinesMistral/Llama3/Phi-2 · MT-Benchscore: 7.8 -> 7.7No clear effect
use adaptive scheduler instead of fixed gamma'=0.1 (ablation)SSPO Mistral-7B 10% · AlpacaEval 2.0 LC (UltraFeedback)LC win rate: 27.5 -> 30.0Improved
use adaptive scheduler instead of fixed gamma'=0.5 (ablation)SSPO Mistral-7B 10% · AlpacaEval 2.0 LC (UltraFeedback)LC win rate: 29.3 -> 30.0Improved
use prior 0.5 instead of 0.1 (ablation)SSPO Mistral-7B 10% · AlpacaEval 2.0 LC (UltraFeedback)LC win rate: 25.6 -> 30.0Improved
use prior 0.5 instead of 0.9 (ablation)SSPO Mistral-7B 10% · AlpacaEval 2.0 LC (UltraFeedback)LC win rate: 25.7 -> 30.0Improved
use SSPO instead of DPO on toy taskGPT-2-small toy, n_L=10 · toy dataset noise 0%test accuracy: 0.743 -> 0.841Improved
use SSPO instead of DPO with 50% label noise and n_L=100GPT-2-small toy · toy dataset noise 50%test accuracy: 0.554 -> 0.563No clear effect

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