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

Consistent One-vs-All Losses Robust to Misspecification of the Weak Label Transition Model

NeurIPS 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Partial and multi-label learning

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes consistent one-vs-all losses for learning from weak labels that remain robust when the weak-label transition model is misspecified.

Problem

Learning from weak labels (noisy, partial, complementary) requires modeling the dependence between true and observed labels, but existing methods rely on transition models that are assumed known, accurately estimated, or instance-independent, and degrade under misspecification.

Approach

Introduces convex losses for weak-label learning built on a one-vs-all formulation, designed to be consistent and robust to misspecification of the weak-label transition model rather than depending on a particular assumed transition model.

Closest papers by meaning

• Losses over Labels: Weakly Supervised Learning via Direct Loss Construction (AAAI 2023 · similarity 0.92)
• Loss factorization, weakly supervised learning and label noise robustness (ICML 2016 · similarity 0.91)
• Learning from weak labelers as constraints (ICLR 2025 · similarity 0.90)
• Multi-Label Learning with Stronger Consistency Guarantees (NeurIPS 2024 · similarity 0.88)
• Forming Auxiliary High-confident Instance-level Loss to Promote Learning from Label Proportions (CVPR 2025 · similarity 0.88)
• Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.87)