Research map / Multi-agent systems and games / Games and equilibria
Neural Payoff Machines: Predicting Fair and Stable Payoff Allocations Among Team Members
Daphne Cornelisse, Thomas Rood, Yoram Bachrach, Mateusz Malinowski, Tal Kachman
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 |
|---|---|---|---|
| Use a variable-size zero-padded model instead of a fixed-size model (ablation) | Neural payoff machines for WVGs · WVG in-sample | Least Core excess MAE: 0.029 -> 0.051 | Hurt |
| Use a variable-size zero-padded model instead of a fixed-size model (ablation) | Neural payoff machines for WVGs · WVG in-sample | Shapley value MAE: 0.018 -> 0.022 | Hurt |
| Use a variable-size zero-padded model instead of a fixed-size model (ablation) | Neural payoff machines for WVGs · WVG in-sample | Banzhaf index MAE: 0.017 -> 0.027 | Hurt |
| Use a variable-size zero-padded model instead of a fixed-size model (ablation) | Neural payoff machines for WVGs · WVG in-sample | Least Core payoff MAE: 0.019 -> 0.015 | Mixed |
| Replace multinomial regression with a neural payoff machine | WVG Shapley estimation · WVG, 4 players | Shapley value MAE: 0.137 -> 0.07 | Improved |
| Replace multinomial regression with a neural payoff machine | WVG Shapley estimation · WVG, 20 players | Shapley value MAE: 0.025 -> 0.006 | Improved |
| Increase training data from 1% to 3% | Melbourne Housing Shapley prediction · Melbourne Housing | RMSE: 19.72 -> less than 0.07 | Improved |
| Increase training data from 0.5% to 4% | UCI Banking Shapley prediction · UCI Banking | RMSE: 4.10 -> 0.25 | Improved |
| Predict the remaining Shapley values with a neural model rather than compute all with SHAP | Melbourne Housing feature importance · Approximately 9000 instances | Speedup: 8x | Improved |
| Use neural prediction instead of SHAP for Shapley value estimation | Melbourne Housing feature importance · Approximately 9000 instances | MSE: 7.4 \times 10^{-5} -> 1.15 \times 10^{-3} | Hurt |
| Evaluate variable-size WVG models on player counts beyond training | Variable-size neural payoff machines · WVGs with more than 10 players | MAE | No clear effect |
| Increase quota across a solution discontinuity | Fixed-size WVG neural payoff machines · Weighted voting games | Solution prediction accuracy | Hurt |
Closest papers by meaning
- Shapley-Coop: Credit Assignment for Emergent Cooperation in Self-Interested LLM Agents (NeurIPS 2025 · similarity 0.86)
- COSAC: Counterfactual Credit Assignment in Sequential Cooperative Teams (NeurIPS 2026 · similarity 0.86)
- Aligning Individual and Collective Objectives in Multi-Agent Cooperation (NeurIPS 2024 · similarity 0.86)
- Approximating the Shapley Value without Marginal Contributions (AAAI 2024 · similarity 0.86)
- Self-Play $Q$-Learners Can Provably Collude in the Iterated Prisoner's Dilemma (ICML 2025 · similarity 0.85)
- Training Generalizable Collaborative Agents via Strategic Risk Aversion (NeurIPS 2026 · 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.
