Research map / Probabilistic and scientific modeling / Molecules and proteins
Game Design for Eliciting Distinguishable Behavior
Fan Yang, Liu Leqi, Yifan Wu, Zachary Lipton, Pradeep K Ravikumar, Tom M. Mitchell, William W. Cohen
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 |
|---|---|---|---|
| optimize reward with mutual information objective instead of manual reward | Path (1 x 6) game · player classification | classification accuracy: 0.442 (0.056) -> 0.678 (0.044) | Improved |
| optimize reward with mutual information objective instead of manual reward | Grid (3 x 6) game · mutual information loss | mutual information loss: 0.115 -> 0.099 | Improved |
| also learn transition probability in addition to reward | Grid (3 x 6) · player classification | classification accuracy: 0.658 (0.066) -> 0.822 (0.027) | Improved |
| also learn transition probability in addition to reward | Path (1 x 6) · player classification | classification accuracy: 0.678 (0.044) -> 0.686 (0.044) | No clear effect |
| also learn transition probability in addition to reward | Grid (3 x 6) · mutual information loss | mutual information loss: 0.099 -> 0.078 | Improved |
| use Grid (3 x 6) instead of Path (1 x 6) when only reward is learned | learned games · mutual information loss | mutual information loss: 0.108 -> 0.099 | Improved |
| use Grid instead of Path when only reward is learned | learned games · player classification | classification accuracy: 0.678 (0.044) -> 0.658 (0.066) | No clear effect |
| increase interaction model noise lambda from 1 to 2.5 | Grid (3 x 6) learned reward and transition · player classification | classification accuracy: 0.822 (0.027) -> 0.730 (0.061) | Hurt |
| increase interaction model noise lambda from 1 to 2.5 | Path (1 x 6) learned reward only · player classification | classification accuracy: 0.678 (0.044) -> 0.650 (0.048) | Hurt |
| increase interaction model noise lambda from 1 to 1.5 | manually designed Path baseline · player classification | classification accuracy: 0.442 (0.056) -> 0.510 (0.053) | Improved |
| replace Full player prior with Diagonal prior (ablation) | learned games (all four settings) · mutual information loss | mutual information loss: 0.078 -> 0.036 | Improved |
| replace Full player prior with Diagonal prior (ablation) | Grid (3 x 6) learned reward and transition · player classification | classification accuracy: 0.822 (0.027) -> 0.712 (0.051) | Hurt |
Closest papers by meaning
- Detecting Individual Decision-Making Style: Exploring Behavioral Stylometry in Chess (NeurIPS 2021 · similarity 0.88)
- Who’s Gaming the System? A Causally-Motivated Approach for Detecting Strategic Adaptation (NeurIPS 2024 · similarity 0.86)
- ElementaryNet: A Non-Strategic Neural Network for Predicting Human Behavior in Normal-Form Games (AAAI 2026 · similarity 0.86)
- Evolution of Information in Interactive Decision Making: A Case Study for Multi-Armed Bandits (NeurIPS 2025 · similarity 0.84)
- Decoding Rewards in Competitive Games: Inverse Game Theory with Entropy Regularization (ICML 2025 · similarity 0.84)
- Modeling Boundedly Rational Agents with Latent Inference Budgets (ICLR 2024 · similarity 0.84)
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.
