AI Paper Atlas

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

GrayKD: Distilling Better Knowledge from Black-box LLM via Multi-rationale Injection

Hyeongsoo Lim, Hyung Yong Kim, Jin Young Kim, Min Ho Jang, Eun Seo Seo, Youshin Lim, Shukjae Choi, Jihwan Park, Yunkyu Lim, Hanbin Lee, Byeong-Yeol Kim, Ji Won Yoon

AAAI 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation with causal and adaptive guidance · 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 mix-sample strategy (ablation)GrayKD (Triple) · 5 benchmarks avgRouge-L: 26.85 -> 27.64Improved
add rationale masking (ablation)GrayKD (Triple) · 5 benchmarks avgRouge-L: 27.46 -> 27.64Improved
add both mix-sample and masking (ablation)GrayKD (Triple) · 5 benchmarks avgRouge-L: 26.34 -> 27.64Improved
add masking (ablation)GrayKD (Triple) · DollyRouge-L: 26.66 -> 26.55No clear effect
use less diverse rationales (Sim 1, cosine 0.8618) (ablation)GrayKD (Triple) · 5 benchmarks avgRouge-L: 27.64 -> 27.29Hurt
reuse a single rationale across all encoders (Sim 2) (ablation)GrayKD (Triple) · 5 benchmarks avgRouge-L: 27.64 -> 26.50Hurt
use three rationales instead of twoGrayKD · 5 benchmarks avgRouge-L: 27.44 -> 27.64Improved
use GrayKD (Triple) instead of PromptKD with White Teacher 3SmolLM2-360M student · 5 benchmarks avgRouge-L: 26.44 -> 27.64Improved
distill with black-box SeqKD instead of white-box KDSmolLM2-360M student · 5 benchmarks avgRouge-L: 26.11 -> 23.59Hurt
add ProxyKD to PromptKD with White Teacher 3SmolLM2-360M student · 5 benchmarks avgRouge-L: 26.44 -> 24.45Hurt
set lambda1=0.4 versus lambda1=0 (distillation only) (ablation)GrayKD · dev setRouge-LImproved
replace 1.7B white-box teacher training with rationale injectionGrayKD · total parameters usedparameters: 2.06B -> 610MImproved

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