Research map / Multi-agent systems and games / Games and equilibria
Opponent modeling in competitive games
Shapes and adapts to opponents in two-player games via auto-curricula and strategy representations.
75 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Games and equilibria, keyword cluster “game · player · sum”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
21 papers in 2023–24 and 10 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 16.5% to 7.0%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Emergence of Social Punishment and Cooperation through Prior Commitments (AAAI 2016 · 32 citations)
- 2017 · Playing FPS Games with Deep Reinforcement Learning (AAAI 2017 · 456 citations)
- 2018 · Re-evaluating evaluation (NeurIPS 2018 · 36 citations)
- 2019 · Diverse Generation for Multi-Agent Sports Games (CVPR 2019 · 104 citations)
- 2020 · Mastering Complex Control in MOBA Games with Deep Reinforcement Learning (AAAI 2020 · 276 citations)
- 2021 · SCC: an efficient deep reinforcement learning agent mastering the game of StarCraft II (ICML 2021 · 30 citations)
- 2022 · Q-Ball: Modeling Basketball Games Using Deep Reinforcement Learning (AAAI 2022 · 19 citations)
- 2023 · MAESTRO: Open-Ended Environment Design for Multi-Agent Reinforcement Learning (ICLR 2023 · 3 citations)
- 2024 · Approximating the Shapley Value without Marginal Contributions (AAAI 2024 · 28 citations)
- 2025 · Advantage Alignment Algorithms (ICLR 2025 · 1 citation)
- 2026 · Look-ahead Reasoning with a Learned Model in Imperfect Information Games (ICLR 2026 · 0 citations)
Earliest papers here
- Emergence of Social Punishment and Cooperation through Prior Commitments (AAAI 2016 · 32 citations)
- Efficient Computation of Emergent Equilibrium in Agent-Based Simulation (AAAI 2016 · 1 citation)
- Playing FPS Games with Deep Reinforcement Learning (AAAI 2017 · 456 citations)
What has been tried
64 papers here report 504 measured changes (a component added, removed or swapped, compared with a baseline): helped 65%, no clear effect 9%, hurt 20%, mixed 6%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Replace the proposed coalition-size distribution with uniform sampling over sizes 2 to n-2 | Synthetic and explainability games · MSE | No clear effect | Approximating the Shapley Value without Marginal Contributions (AAAI 2024 · 28 citations) | |
| Train the discussion policy with RL instead of prompting the LLM to choose tactics | Three-player ONUW · NashConv | Helped | Learning to Discuss Strategically: A Case Study on One Night Ultimate Werewolf (NeurIPS 2024 · 2 citations) | |
| Replace the discount-based regularizer with a $1/(1+t)$ regularizer | Advantage Alignment implementation · performance | Hurt | Advantage Alignment Algorithms (ICLR 2025 · 1 citation) | |
| Mask class names and remove comments and docstrings | SPARC Masked · accuracy | No clear effect | Evaluating LLMs in Open-Source Games (NeurIPS 2025 · 0 citations) | |
| Add the exploration reward | PO-Overcooked · Success rate | 0.173 ± 0.016 -> 0.553 ± 0.029 | Helped | Fast Peer Adaptation with Context-aware Exploration (ICML 2024 · 0 citations) |
| Replace the standard MCTS with fixed-depth search and Logit-equilibrium backup | Simultaneous games · performance | Helped | Mastering Zero-Shot Interactions in Cooperative and Competitive Simultaneous Games (ICML 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 7 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- PPO · 2 of 7 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 31 papers; from OpenAlex and the papers' affiliations).
- Peking University · 3 papers
- Chinese Academy of Sciences · 2 papers
- University of California, Berkeley · 2 papers
- University College London · 2 papers
- Google DeepMind · 2 papers
- University of Oxford · 2 papers
- Meta · 2 papers
- Carnegie Mellon University · 2 papers
- The Chinese University of Hong Kong · 2 papers
Use it
Download BibTeX of 39 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Learning in Markov games (111 papers)
- Cooperation and social dilemmas (95 papers)
- Mean-field game learning (21 papers)
Elsewhere in the topic
- Fair division and coalition formation (99 papers)
- Learning with strategic agents (74 papers)
- Strategic classification (51 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).