Research map / Learning theory and algorithms / Submodular and game algorithms
Responsible AI (RAI) Games and Ensembles
Yash Gupta, Runtian Zhai, Arun Suggala, Pradeep Kumar Ravikumar
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 ERM with RAI-FW (chi2) ensemble, domain oblivious | small base learners · CIFAR10 | worst-class loss: 27.0 -> 23.4 | Improved |
| replace ERM with RAI-FW (chi2) ensemble, domain oblivious | small base learners · CIFAR100 | worst-class loss: 66.0 -> 63.8 | Improved |
| replace Online GDRO with RAI-FW (Group), domain aware | small base learners · CIFAR100 | worst-class loss: 57.0 -> 52.9 | Improved |
| replace Online GDRO with RAI-FW (Group), domain aware | small base learners · CIFAR10 | worst-class loss: 21.7 -> 20.3 | Improved |
| replace Online GDRO with RAI-FW (Group and chi2), partial domain aware | small base learners · CIFAR100 | worst-class loss: 64.0 -> 58.9 | Improved |
| replace Online GDRO with RAI-GA (Group and chi2), partial domain aware | small base learners · COMPAS | worst-group loss: 32.7 -> 32.9 | No clear effect |
| use AdaBoost instead of ERM | small base learners · CIFAR-10 (Imbalanced) | worst-class loss: 33.6 -> 40.6 | Hurt |
| use RAI-FW (chi2) ensemble instead of ERM, average-case check | small base learners · CIFAR100 | average loss: 27.1 -> 27.5 | No clear effect |
| use RAI-FW (Group) instead of Online GDRO, average-case check | small base learners · COMPAS | average loss: 31.7 -> 32.1 | No clear effect |
| use complex larger base models with RAI-GA instead of ERM complex | complex models · CIFAR10 | worst-class loss: 21.3 -> 15.4 | Improved |
| use larger complex models instead of small models with RAI ensemble | RAI-FW (chi2) · CIFAR10 | worst-class loss: 23.4 -> 15.4 | Improved |
| use RAI-GA instead of RAI-FW (chi2) on small DO | small base learners · CIFAR10 | worst-class loss: 25.0 -> 23.4 | No clear effect |
Closest papers by meaning
- Backward Responsibility in Transition Systems Using General Power Indices (AAAI 2024 · similarity 0.85)
- Estimating α-Rank by Maximizing Information Gain (AAAI 2021 · similarity 0.85)
- Learning Diverse Risk Preferences in Population-Based Self-Play (AAAI 2024 · similarity 0.83)
- Playing Markov Games Without Observing Payoffs (NeurIPS 2026 · similarity 0.83)
- Improved Regret Bounds for Online Fair Division with Bandit Learning (AAAI 2025 · similarity 0.83)
- The Complexity of Computing Robust Mediated Equilibria in Ordinal Games (AAAI 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.
