Research map / Label-efficient and robust learning / Recognition and long-tail
Order-Free Learning Alleviating Exposure Bias in Multi-Label Classification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| train the RNN decoder with OCD on sampled model-generated prefixes instead of correct target prefixes | RNN-based multi-label classification · AAPD resplit, unseen test set | ebF1: 0.515 -> 0.530 | Helped |
| train only with correct prefixes (ablation) | OCD · AAPD resplit, unseen test set | ebF1: 0.530 -> 0.515 | Hurt |
| add the auxiliary binary-relevance decoder through multitask learning | OCD · AAPD | maF1: 0.541 -> 0.578 | Helped |
| combine BR and RNN probabilities using logistic rescoring | OCD + MTL · AAPD | Average: 0.676 -> 0.682 | Helped |
| use logistic joint decoding instead of logistic rescoring | OCD + MTL · Reuters-21758 | Average: 0.819 -> 0.822 | Helped |
| use logistic joint decoding instead of logistic rescoring | OCD + MTL · Audio set | Average: 0.482 -> 0.487 | Helped |
| use logistic joint decoding instead of logistic rescoring | OCD + MTL · RCV1-V2 | Average: 0.815 -> 0.816 | Helped |
| add scheduled sampling to Seq2seq | Seq2seq · AAPD resplit, unseen test set | ebF1: 0.508 -> 0.517 | Helped |
| add scheduled sampling to Order-free RNN | Order-free RNN · AAPD resplit, unseen test set | ebF1: 0.496 -> 0.520 | Helped |
| add scheduled sampling | Seq2seq · AAPD | Average: 0.662 -> 0.667 | Mixed |
| add scheduled sampling | Order-free RNN · RCV1-V2 | Average: 0.809 -> 0.806 | Hurt |
| remove the predefined target order and train with OCD | RNN-based multi-label classification · AAPD | unseen generated label combinations: 1 -> 30 | Helped |
Closest papers by meaning
- Enhancing Multi-Label Classification via Dynamic Label-Order Learning (AAAI 2024 · similarity 0.89)
- Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection (AAAI 2025 · similarity 0.86)
- Rebalancing Multi-Label Class-Incremental Learning (AAAI 2025 · similarity 0.84)
- Generalization Analysis for Multi-Label Learning (ICML 2024 · similarity 0.84)
- Batch Selection for Multi-Label Classification Guided by Uncertainty and Dynamic Label Correlations (AAAI 2025 · similarity 0.84)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.84)
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.