Research map / Label-efficient and robust learning / Recognition and long-tail
When More Experts Hurt: Underfitting in Multi-Expert Learning to Defer
Shuqi Liu, Yuzhou Cao, Lei Feng, Bo An, Luke Ong
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace vanilla CE surrogate with PiCCE-CE, 20 dog experts | multi-expert CE L2D · ImageNet Dog Expert, 20 experts | system error: 48.00 -> 41.32 | Improved |
| replace vanilla OvA surrogate with PiCCE-OvA, 20 dog experts | multi-expert OvA L2D · ImageNet Dog Expert, 20 experts | system error: 66.56 -> 42.25 | Improved |
| replace vanilla CE with PiCCE-CE, 20 animal experts | multi-expert CE L2D · CIFAR-100 Animal Expert, 20 experts | system error: 21.13 -> 18.11 | Improved |
| replace vanilla OvA with PiCCE-OvA, 20 animal experts | multi-expert OvA L2D · CIFAR-100 Animal Expert, 20 experts | system error: 22.91 -> 18.63 | Improved |
| PiCCE-OvA coverage with 20 animal experts | multi-expert OvA L2D · CIFAR-100 Animal Expert, 20 experts | coverage: 77.71 -> 86.69 | Improved |
| PiCCE-CE with 8 MiceBone experts | multi-expert CE L2D · MiceBone, 8 experts | system error: 16.72 -> 13.35 | Improved |
| PiCCE-OvA with 8 Chaoyang experts | multi-expert OvA L2D · Chaoyang, 8 experts | system error: 2.48 -> 1.02 | Improved |
| PiCCE-CE with 2 MiceBone experts | multi-expert CE L2D · MiceBone, 2 experts | system error: 15.17 -> 15.23 | No clear effect |
| add more experts to vanilla CE | multi-expert CE L2D · CIFAR-100 Animal Expert | system error: 18.48 -> 21.13 | Hurt |
| add more experts to PiCCE-CE | PiCCE-CE · CIFAR-100 Animal Expert | system error: 18.32 -> 18.11 | Improved |
| use a hard argmax pick of the confident expert instead of PiCCE (ablation) | Pick the Confident Expert loss · multi-expert L2D | surrogate continuity | Hurt |
Closest papers by meaning
- Regression with Multi-Expert Deferral (ICML 2024 · similarity 0.87)
- Mastering Multiple-Expert Routing: Realizable $H$-Consistency and Strong Guarantees for Learning to Defer (ICML 2025 · similarity 0.85)
- A Unifying Post-Processing Framework for Multi-Objective Learn-to-Defer Problems (NeurIPS 2024 · similarity 0.84)
- Exploiting Human-AI Dependence for Learning to Defer (ICML 2024 · similarity 0.84)
- When to Act and When to Ask: Policy Learning With Deferral Under Hidden Confounding (NeurIPS 2024 · similarity 0.83)
- LFME: A Simple Framework for Learning from Multiple Experts in Domain Generalization (NeurIPS 2024 · 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.
