Research map / Label-efficient and robust learning / Recognition and long-tail
Loss Functions for Multiset Prediction
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 |
|---|---|---|---|
| Use the multiset loss rather than policy-gradient RL | multiset prediction · MNIST Multi (10) | EM: 0.665 -> 0.920 | Helped |
| Use the multiset loss rather than policy-gradient RL | multiset prediction · MNIST Multi (4) | F1: 0.977 -> 0.987 | Helped |
| Use the multiset loss rather than sequence loss | multiset prediction · MNIST Multi (1-4) | EM: 0.891 -> 0.953 | Helped |
| Use the multiset loss rather than one-step distribution matching | multiset prediction · MNIST Multi (4) | EM: 0.210 -> 0.950 | Helped |
| Use multiset loss rather than aggregated KL distribution matching | multiset prediction · COCO Medium | EM: 3.7 percentage points | Helped |
| Use multiset loss rather than aggregated KL distribution matching | multiset prediction · COCO Medium | F1: 4.8 | Helped |
| Use a random rank rather than an area-based rank | sequence loss · MNIST Multi (4) | EM: 0.529 -> 0.920 | Helped |
| Use a random rank rather than a spatial rank | sequence loss · COCO Easy | EM: 0.675 -> 0.721 | Helped |
| Train with oracle sampling rather than policy-visited states (ablation) | multiset loss · COCO Medium | F1: 0.645 -> 0.616 | Hurt |
| Train with stochastic sampling rather than greedy decoding (ablation) | multiset loss · COCO Medium | EM: 0.475 $\pm$ 0.006 vs. 0.475 $\pm$ 0.004 | No clear effect |
| Train with stochastic sampling rather than greedy decoding (ablation) | multiset loss · COCO Medium | F1: 0.645 $\pm$ 0.016 vs. 0.649 $\pm$ 0.009 | No clear effect |
| Train with oracle sampling rather than policy-visited states (ablation) | multiset loss · COCO Medium | EM: 0.475 $\pm$ 0.006 -> 0.469 $\pm$ 0.002 | Hurt |
Closest papers by meaning
- Addressing the Loss-Metric Mismatch with Adaptive Loss Alignment (ICML 2019 · similarity 0.86)
- Leaving the Nest: Going beyond Local Loss Functions for Predict-Then-Optimize (AAAI 2024 · similarity 0.86)
- Toward a Characterization of Loss Functions for Distribution Learning (NeurIPS 2019 · similarity 0.85)
- Robust Loss Functions for Training Decision Trees with Noisy Labels (AAAI 2024 · similarity 0.85)
- A Survey of Learning Criteria Going beyond the Usual Risk (Abstract Reprint) (AAAI 2024 · similarity 0.83)
- Loss Functions and Operators Generated by f-Divergences (ICML 2025 · similarity 0.83)
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.