AI Paper Atlas

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

Don't Ignore the Tail: Decoupled Distillation Produces Top Maths Students on an Academic Budget

Sayantan Dasgupta, Trevor Cohn, Timothy Baldwin

ICML 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives

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
replace KL with tail-aware divergence (K=10)pretraining distillation, Qwen1.5-1.8B -> 1.2B · 8-task avg Regmixaverage accuracy (%): 45.6 -> 47.8Improved
replace KL with tail-aware divergence (K=10)Phi2 2.8B -> 1.1B · 8-task avg Regmixaverage accuracy (%): 49.1 -> 50.3Improved
replace KL with tail-aware divergence (K=10)Gemma2-9B -> 2B · 8-task avg Regmixaverage accuracy (%): 45.8 -> 47.0Improved
replace KL with tail-aware divergence (K=10)Qwen2.5-3B -> 1.2B · 8-task avg Regmixaverage accuracy (%): 44.9 -> 46.3Improved
use MiniPLM data selection instead of vanilla KDPhi2 2.8B -> 1.1B · 8-task avg Regmixaverage accuracy (%): 49.1 -> 48.8Hurt
increase top-K beyond 10 (ablation)TAD · 8-task avg Regmixaverage accuracy (%): 47.8 -> 47.7No clear effect
vary beta (0.5 to 10) with K=10 (ablation)TAD, Qwen1.8B -> 1.2B · 8-task avg Regmixaverage accuracy (%): 47.0 / 47.6 / 47.8 / 47.7 / 47.6No clear effect
add cosine loss on hidden statesVanilla KD and TAD, larger teachers · 8-task avg Regmixaverage accuracy (%)Improved
use reverse KL instead of forward KL (ablation)distillation with cosine loss · Full-ECE RegmixF-ECE: 1.45 -> 1.77Hurt
use TAD instead of vanilla KDPhi2 2.8B -> 1.1B · Full-ECE RegmixF-ECE: 1.45 -> 1.19Improved
increase K from 1 to 20 (ablation)TAD · Full-ECEF-ECE: 1.19 -> 1.42Hurt
use MiniPLM for domain-specific distillationTinyLlama-1.1B on OWM · math 8-task avgaverage accuracy (%): 16.7 -> 14.4Hurt

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