Research map / Learning theory and algorithms / Prediction and decision losses
Delegated Classification
Eden Saig, Inbal Talgam-Cohen, Nir Rosenfeld
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 |
|---|---|---|---|
| increase validation set size m | min-budget contracts · MNIST, target accuracy 0.85 | required budget B* | Improved |
| switch from MLP to GBDT as delegated learner at accuracy ~0.74 | delegated learning · MNIST | required budget: DeltaB* = -10^2 | Improved |
| switch from MLP to GBDT as delegated learner at accuracy ~0.94 | delegated learning · MNIST | required budget: DeltaB* = 10^5 | Hurt |
| restrict optimization to simple contracts instead of general min-budget | contract design · LCDB multiple learners | required budget: less than 1% higher | No clear effect |
| try SBA before generic LP solver | min-budget contract solver · LCDB | solve time and success: more than 85% success, ~10^3 times faster | Improved |
| low bias (larger n0, r=1) instead of low variance (more samples per n) for curve fit | learning curve extrapolation · MNIST partial information | accuracy via delegation | Improved |
| increase pilot set size k | learning curve estimation · MNIST partial information | accuracy obtained via delegation | Improved |
| use r=1 rather than r>1 samples per n | sample-size multiplier · MNIST partial information | mu(k): ~10 -> ~30 | Improved |
| over-estimate learning curve instead of under-estimate | contract from extrapolated curve · MNIST partial information | accuracy loss from delegation | Hurt |
| under-estimate learning curve | contract from extrapolated curve · MNIST partial information | accuracy loss from delegation | No clear effect |
| relax C-MLRP assumption for threshold contracts | threshold contracts · LCDB | optimality of threshold form | No clear effect |
Closest papers by meaning
- Paid with Models: Optimal Contract Design for Collaborative Machine Learning (AAAI 2025 · similarity 0.89)
- Contracting with a Learning Agent (NeurIPS 2024 · similarity 0.86)
- Learning Optimal Contracts: How to Exploit Small Action Spaces (ICLR 2024 · similarity 0.86)
- Bayesian Strategic Classification (NeurIPS 2024 · similarity 0.85)
- Contract Design Under Approximate Best Responses (ICML 2025 · similarity 0.85)
- Strategic Classification With Externalities (ICLR 2025 · similarity 0.85)
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.
