Research map / Learning theory and algorithms / Prediction and decision losses
Multiclass Performance Metric Elicitation
Gaurush Hiranandani, Shant Boodaghians, Ruta Mehta, Oluwasanmi O Koyejo
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 dataset size from 50% to 75% to 100% of datapoints | DLPM elicitation (Algorithm 1) with estimated confusions · SensIT (Acoustic) and Vehicle | proportion of successful elicitation within omega | Improved |
| use the full SensIT (Acoustic) dataset | Algorithm 1 · SensIT (Acoustic) | omega for eliciting all metrics: 0.12 | Improved |
| use a very tight evaluation tolerance omega in [0.04, 0.08] | Algorithm 1 · SensIT and Vehicle | successful ME proportion: [0.04, 0.08] | Hurt |
| set eps = 0.01 with synthetic oracle | Algorithm 1 (DLPM) · Synthetic k=3, k=4 | query count | No clear effect |
| increase classes from k=3 to k=4 | Algorithm 2 (LPM) · Synthetic | number of queries: 320 -> 704 | Hurt |
| increase classes from k=3 to k=4 | Algorithm 1 (DLPM) · Synthetic | number of queries: 56 -> 84 | Hurt |
| use two cycles (T = 2(q-1)) of coordinate updates | Algorithm 2 · Synthetic LPM, eps=10^-2 | elicitation error | Improved |
Closest papers by meaning
- A General Online Algorithm for Optimizing Complex Performance Metrics (ICML 2024 · similarity 0.83)
- Foundations of the Theory of Performance-Based Ranking (CVPR 2025 · similarity 0.82)
- Principled Algorithms for Optimizing Generalized Metrics in Binary Classification (ICML 2025 · similarity 0.81)
- Overcoming Common Flaws in the Evaluation of Selective Classification Systems (NeurIPS 2024 · similarity 0.81)
- The Implicit Bias of Gradient Descent on Separable Multiclass Data (NeurIPS 2024 · similarity 0.81)
- Preference learning made easy: Everything should be understood through win rate (ICML 2025 · similarity 0.80)
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.
