AI Paper Atlas

Research map / Large language models / Alignment and preferences

Alignment through Meta-Weighted Online Sampling: Bridging the Gap between Data Generation and Preference Optimization

Junming Yang, Ning Xu, Biao Liu, Shiqi Qiao, Xin Geng

ICLR 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO · 3 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 meta-weighted adaptive sampling instead of random sampling (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2LC win rate (%): 38.53 -> 43.21Improved
add meta-weighted adaptive sampling instead of threshold sampling (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2LC win rate (%): 42.92 -> 43.21Improved
add meta-weighted selective sampling instead of sampling all prompts online (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2LC win rate (%): 42.04 -> 43.21Improved
add meta-weighted selective sampling instead of sampling all prompts online (ablation)MetaAPO (Llama-3.1-8B, DPO) · Annotation ratioonline annotation ratio (%): 100.0 -> 58.4Improved
add learned meta-weights for loss instead of uniform 0.5 weighting (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2LC win rate (%): 37.70 -> 43.21Improved
add learned meta-weights for loss instead of uniform 0.5 weighting (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2raw win rate (%): 39.25 -> 47.48Improved
add embedding similarity of chosen/rejected as meta-learner input (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2LC win rate (%): 43.21 -> 42.88No clear effect
add continual meta-learner updates instead of freezing the learned meta-learner (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2LC win rate (%): 40.03 -> 43.21Improved
add trainable meta-learner instead of fixed sigmoid heuristic (ablation)MetaAPO (Llama-3.1-8B, DPO) · AlpacaEval 2raw win rate (%): 43.07 -> 47.48Improved
add trainable meta-learner instead of fixed sigmoid heuristic (ablation)MetaAPO (Llama-3.1-8B, DPO) · Arena-Hardwin rate (%): 38.3 -> 43.9Improved
add MetaAPO on top of Online DPOOnline DPO (Llama-3.1-8B) · AlpacaEval 2LC win rate (%): 39.98 -> 43.21Improved
MetaAPO versus PPO with fewer online samplesPPO (Llama-3.1-8B) · Arena-Hardwin rate (%): 40.8 -> 43.9Improved

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