AI Paper Atlas

Research map / Large language models / Alignment and preferences

Learning from Noisy Preferences: A Semi-Supervised Learning Approach to Direct Preference Optimization

Xinxin Liu, Ming Li, Zonglin Lyu, Yuzhang Shang, Chen Chen

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 first self-training iteration of pseudo-labels on noisy data after clean cold-start (ablation)Semi-DPO (SD1.5) · HPS v2 promptsImageReward: 0.569 -> 0.798Improved
add second self-training iteration (ablation)Semi-DPO (SD1.5) · HPS v2 promptsImageReward: 0.798 -> 0.816Improved
add second self-training iteration (ablation)Semi-DPO (SD1.5) · HPS v2 promptsCLIP: 0.300 -> 0.299No clear effect
add second iteration (training cost) (ablation)Semi-DPO · training costGPU hours: 132 -> 228Hurt
enlarge consensus from 2 to 5 reward models (ablation)Semi-DPO Iter0 (SD1.5) · HPS v2 promptsImageReward: 0.442 -> 0.569Improved
enlarge consensus from 2 to 5 reward models (ablation)Semi-DPO Iter0 (SD1.5) · HPS v2 promptsMPS: 12.891 -> 13.039Improved
enlarge consensus from 2 to 5 reward models (ablation)Semi-DPO Iter0 (SD1.5) · Pick-a-Pic V2 promptsImageReward: 0.417 -> 0.563Improved
use dynamic per-timestep thresholds to filter pseudo-labels (vs full Diffusion-DPO) (ablation)Semi-DPO · training costGPU hours: 192 -> 132Improved
replace Diff-DPO with Semi-DPO (ablation)SD1.5 · GenEvaloverall score: 43.00 -> 47.31Improved
replace Diff-DPO with Semi-DPO (ablation)SDXL · GenEvaloverall score: 58.02 -> 58.41Improved
replace Diff-DPO with Semi-DPO (ablation)SDXL · HPS v2 promptsPickScore: 22.953 -> 22.816Hurt
replace Diff-KTO with Semi-DPO (ablation)SD1.5 · HPS v2 promptsImageReward: 0.690 -> 0.816Improved

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