AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

HGLTR: Hierarchical Knowledge Injection for Calibrating Pre-trained Models in Long-Tail Recognition

Jinpeng Zheng, Shao-Yuan Li, Gan Xu, Wenhai Wan, Zijian Tao, Songcan Chen, Kangkan Wang

AAAI 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Decoupled and calibrated long-tail training · 1 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 Ldis (hierarchical distillation) to Lmse (ablation)HGLTR · ImageNet-LTtail accuracy: 68.5 -> 72.1Improved
add Lmse (feature alignment) to Ldis (ablation)HGLTR · ImageNet-LTtail accuracy: 69.5 -> 72.1Improved
add both hierarchical losses (ablation)HGLTR · ImageNet-LToverall accuracy: 76.3 -> 77.4 (Lmse only -> both)Improved
use Adapter instead of LoRA/VPT (ablation)HGLTR · ImageNet-LToverall accuracy: 75.7 -> 77.4 (LoRA -> Adapter)Improved
HGLTR vs BALLAD with fewer paramsCLIP ViT-B/16 · ImageNet-LToverall accuracy: 75.7 -> 77.4Improved
HGLTR vs GML using extra dataCLIP ViT-B/16 · ImageNet-LToverall accuracy: 78.0 vs 77.4Hurt
HGLTR vs LPT on iNaturalist tailCLIP ViT-B/16 · iNaturalist 2018tail accuracy: 79.3 vs 78.8Hurt
HGLTR on Places-LTCLIP ViT-B/16 · Places-LToverall accuracy: 50.1 -> 51.4 (vs LPT/GNM-PT)Improved
vary alpha (feature loss weight) (ablation)HGLTR · ImageNet-LToverall accuracyNo clear effect
vary beta (distillation weight) (ablation)HGLTR · ImageNet-LToverall accuracyMixed

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