AI Paper Atlas

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

Let the Prototype Guide You: Robust Aggregation of Sparse Multi-Class Annotations via Annotator Prototype Learning

Ju Chen, Jun Feng, Shenyu Zhang

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data

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 per-annotator confusion matrices with annotator-specific mixtures over class-specific prototypesDS / truth inference · 10 crowdsourcing datasets (avg)accuracy (%): 68.73 -> 74.11Improved
make prototypes class-specific (CPBCC instead of PTBCC) (ablation)PTBCC · 10 crowdsourcing datasets (avg)accuracy (%): 72.40 -> 74.11Improved
make prototypes class-specific (CPBCC instead of PTBCC) (ablation)PTBCC · Val5accuracy (%): 56.00 -> 50.00Hurt
use CPBCC instead of DS on Val7DS · Val7accuracy (%): 22.00 -> 48.00Improved
increase diagonal prior b from 3 to 5 (ablation)CPBCC · Val7accuracy (%): 34.00 -> 48.00Improved
vary d from 0.15 to 0.45 (ablation)CPBCC · 10 datasets (avg)accuracy (%): 73.80 -> 74.11No clear effect
increase number of prototypes S from 4 to 10 (random init) (ablation)CPBCC · 10 datasets (avg)accuracy (%): 71.51 -> 70.24Hurt
increase number of prototypes S from 3 to 8 (random init) (ablation)PTBCC · 10 datasets (avg)accuracy (%): 69.94 -> 68.95Hurt
increase PTBCC initial diagonal strength f from 2 to 5 (ablation)PTBCC · 10 datasets (avg)accuracy (%): 71.12 -> 72.40Improved
use CPBCC instead of GOVERN on AdultGOVERN · Adultaccuracy (%): 69.52 -> 77.78Improved
use CPBCC instead of IBCC on WebIBCC · Webaccuracy (%): 70.41 -> 85.11Improved

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