AI Paper Atlas

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

The Bridge-Garden Dilemma in LLM Distillation: Why Mixing Hard and Soft Labels Works

Guanghui Wang, Kaiwen Lv Kacuila, Zhiyong Yang, Zitai Wang, Jin-Wen Wu, Longtao Huang, Qianqian Xu, Qingming Huang

ICML 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 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
mix hard and soft labels with static weighting vs best pure KDQwen2.5-7B -> 3B distillation · general reasoning avgaccuracy avg: +1.69Improved
use hybrid (risk-guided) rather than best non-hybrid baselineForward KL distillation · Qwen2.5-3B general reasoningaccuracy avg: +1.95Improved
use soft KD instead of hard KDLlama3.2-1B student · general reasoning avgaccuracy avg: 16.42 -> 23.25Improved
use hard KD instead of soft KDQwen2.5-3B student · general reasoning avgaccuracy avg: 51.87 -> 52.45Improved
hybrid hard-soft KD instead of on-policy KDQwen2.5-7B -> 3B · average accuracyaccuracy: +1.43% gain, 9.7x cheaperImproved
add risk-guided hybrid over static weightingLlama3.2-1B student · general reasoning avgaccuracy avg: 24.23 -> 26.27Improved
add curriculum scheduling over static weightingLlama3.2-1B student · general reasoning avgaccuracy avg: 24.23 -> 25.67Improved
add confidence-based weighting over static weightingQwen2.5-3B student · ThmQAaccuracy: 25.45 -> 22.78Hurt
use intermediate mixing weight rather than pure hard or softHybrid KD · multiple tasksaccuracyImproved

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