Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Diverse Conventions for Human-AI Collaboration
Bidipta Sarkar, Andy Shih, Dorsa Sadigh
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 |
|---|---|---|---|
| add mixed-play (beta=0.5) to pure cross-play minimization (ablation) | CoMeDi · Balance Beam | handshake fraction: 0.16 -> 0.00 | Improved |
| add mixed-play (beta=0.5) to pure cross-play minimization (ablation) | CoMeDi · Balance Beam | self-play score: 1.616 -> 2.000 | Improved |
| add mixed-play (beta=0.5) to pure cross-play minimization (ablation) | CoMeDi · Balance Beam | score with left-biased agent: 1.128 -> 1.384 | Improved |
| raise mixed-play weight from 0.5 to 1.0 (ablation) | CoMeDi · Balance Beam | perfect cross-play fraction: 0.00 -> 0.24 | Hurt |
| raise mixed-play weight from 0.5 to 1.0 (ablation) | CoMeDi · Balance Beam | cross-play score: 0.200 -> 0.632 | Hurt |
| mixed-play beta=0.25 instead of 0.0 (ablation) | CoMeDi · Balance Beam | handshake fraction: 0.16 -> 0.00 | Improved |
| cross-play regularization (alpha=1.0) for second agent | CoMeDi · Blind Bandits | convergence to G convention | Improved |
| statistical diversity (ADAP) instead of cross-play | ADAP · Blind Bandits | convergence to G convention | Hurt |
| CoMeDi instead of SP agent | Overcooked agent · Coordination Ring user study | average score: 2.08 -> 3.68 | Improved |
| CoMeDi (beta=1.0) instead of XP (beta=0.0) convention-aware agent (ablation) | Overcooked agent · Coordination Ring user study | average score: 2.48 -> 3.68 | Improved |
| CoMeDi instead of ADAP training population | Overcooked agent · Coordination Ring self-play | average dishes served: 1.90 -> 5.36 | Improved |
| XP (pure cross-play) instead of SP | Overcooked agent · Coordination Ring self-play | average dishes served: 3.47 -> 3.06 | Hurt |
Closest papers by meaning
- Cross-environment Cooperation Enables Zero-shot Multi-agent Coordination (ICML 2025 · similarity 0.88)
- Beyond Single Stationary Policies: Meta-Task Players as Naturally Superior Collaborators (NeurIPS 2024 · similarity 0.87)
- Melting Pot Contest: Charting the Future of Generalized Cooperative Intelligence (NeurIPS 2024 · similarity 0.87)
- MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs (ICLR 2026 · similarity 0.87)
- OvercookedV2: Rethinking Overcooked for Zero-Shot Coordination (ICLR 2025 · similarity 0.86)
- Robust and Diverse Multi-Agent Learning via Rational Policy Gradient (NeurIPS 2025 · 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.
