Research map / Bandits and online learning / Online convex optimization and games
Next-Token Prediction and Regret Minimization
Mehryar Mohri, Clayton Sanford, Jon Schneider, Kiran Vodrahalli, Yifan Wu
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 |
|---|---|---|---|
| train on robustified (Polya-filtered, prefix-masked) data instead of plain BERNOULLI | NanoDO transformer · 8 ground-truth sequence distributions | regret | Improved |
| robustify BERNOULLI model | NanoDO transformer · in-distribution Ber(1/3)->Ber(2/3) | regret: about -0.16 retained | No clear effect |
| use ROBUST_BERNOULLI instead of POLYAURN to stay near original distribution | NanoDO transformer · TV-distance vs BERNOULLI_1 | TV-distance: 1.0000 -> 0.7602 | Improved |
| use ROBUST_BERNOULLI instead of POLYAURN | NanoDO transformer · Next-token TV vs BERNOULLI_1 | Next-Token TV: 0.1529 -> 0.0199 | Improved |
| use ROBUST_BERNOULLI instead of POLYAURN | NanoDO transformer · Next-token TV vs BERNOULLI_2 | Next-Token TV: 0.1655 -> 0.0299 | Improved |
| mask loss on high-regret prefix of Polya sequences (ablation) | ROBUST_BERNOULLI training · training data | TV-distance | Improved |
| robustification vs seed-to-seed variation | ROBUST_BERNOULLI · full-sequence TV | TV-distance: 0.4193 -> 0.7602 | Hurt |
| use a bounded context length equal to original model | robustification (theory) · bounded context | regret | Hurt |
Closest papers by meaning
- No-Regret Learning Under Adversarial Resource Constraints: A Spending Plan Is All You Need! (NeurIPS 2025 · similarity 0.88)
- Regret Bounds for Adversarial Contextual Bandits with General Function Approximation and Delayed Feedback (NeurIPS 2025 · similarity 0.88)
- High-Dimensional Prediction for Sequential Decision Making (ICML 2025 · similarity 0.87)
- Near-Optimal Regret in Adversarial Kernel Bandits (NeurIPS 2026 · similarity 0.87)
- From Average Sensitivity to Small-Loss Regret Bounds under Random-Order Model (NeurIPS 2026 · similarity 0.86)
- Beating Adversarial Low-Rank MDPs with Unknown Transition and Bandit Feedback (NeurIPS 2024 · similarity 0.86)
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.
