AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

SelecTKD: Selective Token-Weighted Knowledge Distillation for LLMs

Haiduo Huang, Jiangcheng Song, Yadong Zhang, Pengju Ren

CVPR 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 26 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 SelecTKD token weighting on top of DistiLLM-2DistiLLM-2, Qwen2-7B to 1.5B · Instruction following avg WRwin rate: 58.69 -> 59.47Improved
Add SelecTKD on top of vanilla KDKD, Qwen2-7B to 1.5B · Instruction following avg WRwin rate: 41.19 -> 43.33Improved
Add SelecTKD on top of SKDSKD, Mistral-7B to Danube2-1.8B · Evol-Instruct WRwin rate: 38.87 -> 38.96No clear effect
Add SelecTKD on top of DistiLLM-2 for mathDistiLLM-2, Qwen2-Math · GSM8KPass@1: 76.27 -> 77.82Improved
Add SelecTKD on top of DistiLLM-2 for codeDistiLLM-2, DS-Coder · HumanEvalPass@1: 59.92 -> 61.72Improved
Distill a VLM with SelecTKDInternVL2-2B · Open VLM Leaderboard averageaverage score: 54.0 -> 57.9Improved
Use Top-k/Spec-k verification instead of Hellinger-based weighting (ablation)SelecTKD · Instruction following avg WRwin rate: 41.70 -> 43.33Improved
Use non-greedy Spec-k instead of greedy Top-k (ablation)SelecTKD · Instruction following avg WRwin rate: 43.09 -> 43.33Improved
Give rejected tokens small weight beta=0.01 instead of masking (ablation)SelecTKD Spec-k · Instruction following avg WRwin rate: 42.83 -> 43.33Improved
Increase k from 5 to 10 or 3/7 (ablation)SelecTKD Spec-k · Instruction following avg WRwin rate: 43.33 -> 43.13No clear effect
Increase rejected-token weight beta to larger values (ablation)SelecTKD · Instruction following avg WRwin rate: 43.33 -> 43.20 (beta 0.1)Hurt
Vary loss function (KL/RKL/SKL/SRKL) inside SelecTKD (ablation)SelecTKD · ROUGE-L on mixed dataROUGE-LNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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