AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Regret Pre-training: Bridging Prior and Posterior Views for Enhanced Knowledge Grounding

Mingkuan Zhao, Xiayu Sun, Wentao Hu, Suquan Chen, Jiaxuan Li, Xiaoyan Zhu, Xin Lai, Jiayin Wang

ICML 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 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
Add global future-conditioned teacher regret loss (alpha=1.0)OLMoE-1B-7B causal LM · BoolQ (zero-shot)accuracy (%): 42.9 -> 61.0Improved
Add local one-token-lookahead teacher regret lossOLMoE-1B-7B causal LM · BoolQ (zero-shot)accuracy (%): 42.9 -> 48.5Improved
Add global regret lossOLMoE-1B-7B causal LM · 7-task averageaccuracy (%): 30.2 -> 33.9Improved
Add local regret lossOLMoE-1B-7B causal LM · 7-task averageaccuracy (%): 30.2 -> 32.2Improved
Use global teacher instead of local teacher (ablation)Regret Pre-training · ARC-Challengeaccuracy (%): 23.4 -> 25.8Improved
Use local teacher instead of global teacher (ablation)Regret Pre-training · CommonsenseQAaccuracy (%): 23.4 -> 26.5Improved
Use local teacher instead of global teacher (ablation)Regret Pre-training · HellaSwagaccuracy (%): 25.1 -> 27.5Improved
Use local teacher instead of global teacher (ablation)Regret Pre-training · PIQAaccuracy (%): 57.2 -> 58.9Improved
Use local vs global teacher (ablation)Regret Pre-training · WinoGrandeaccuracy (%): 50.6 vs 50.8No clear effect
Set regret weight alpha=0.1 instead of 1.0 (ablation)GlobalRegret · downstream tasksaccuracy improvement over baselineNo clear effect
Increase regret weight alpha to 2.0 (ablation)GlobalRegret · downstream tasksaccuracy improvement over baselineHurt
Use alpha=1.0 regret weight (ablation)GlobalRegret · downstream tasksmedian improvement over baselineImproved

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