AI Paper Atlas

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

Holistic Optimal Label Selection for Robust Prompt Learning under Partial Labels

Yaqi Zhao, Haoliang Sun, Yating Wang, Yongshun Gong, Yilong Yin

ECCV 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Partial and multi-label learning · 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
use HopS instead of best baseline at extreme confusionCoOp uni-prompt · Caltech (rand, gamma_c=0.90)accuracy (%): 87.1 -> 89.5Improved
use HopS instead of SCE under severe confusionCoOp uni-prompt · Food (rand, gamma_c=0.90)accuracy (%): 30.8 -> 65.0Improved
use HopS instead of CCCoOp uni-prompt · OxfordPets (rand, gamma_c=0.90)accuracy (%): 74.2 -> 80.5Improved
use HopS instead of CCCoOp uni-prompt · FGVCAircraft (rand, gamma_c=0.90)accuracy (%): 20.4 -> 19.3Hurt
use HopS instead of RCCoOp cls-prompts · OxfordPets (rand, gamma_c=0.90)accuracy (%): 76.3 -> 70.2Hurt
use HopS instead of CroSel (trained on full data)HopS 16-shot · Caltech (insd, gamma_c=0.80)accuracy (%): 11.8 -> 38.9Improved
use HopS instead of Papi (trained on full data)HopS 16-shot · Caltech (rand, gamma_c=0.80)accuracy (%): 52.2 -> 89.5Improved
use HopS instead of LWC under instance-dependent confusionCoOp uni-prompt · Flowers (insd, R18, gamma_c=0.80)accuracy (%): 28.5 -> 69.9Improved
use HopS instead of CCCoOp uni-prompt · Caltech (insd, CLIP candidates, gamma_c=0.80)accuracy (%): 46.2 -> 45.0Hurt
add local (LDF) and global (GOP) modules jointly instead of stand-alone (ablation)HopS · Food (uni-prompt, gamma_c 0.67 and 0.80)GT label identification accuracyImproved
balance loss weight lambda between local and global signals (0.5, 1.0, 2.0) (ablation)HopS · eight datasetsaccuracy (%)Improved
use neighbor-based refinement instead of confidence-based or graph-based (ablation)LDF · partial-label datasetsaccuracy (%)Improved

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