Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Emergence of Grounded Compositional Language in Multi-Agent Populations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Enable verbal communication | Multi-agent policy · Grounded environment, train | physical reward: -0.919 -> -0.332 | Helped |
| Enable verbal communication | Multi-agent policy · Grounded environment, test | physical reward: -0.920 -> -0.392 | Helped |
| Remove verbal symbol utterances while allowing agents to observe other agents' gaze (ablation) | Multi-agent policy · Grounded environment with gaze observation | communication strategy | Mixed |
| Remove verbal symbol utterances and make other agents' gaze unobservable (ablation) | Multi-agent policy · Grounded environment without gaze observation | communication strategy | Mixed |
| Remove verbal and visual observation by the goal recipient (ablation) | Multi-agent policy · Grounded environment without recipient observation | communication strategy | Mixed |
| Use a smaller maximum symbol vocabulary instead of the soft vocabulary-size penalty (ablation) | Communication policy · Compositional language formation | optimization outcome | Hurt |
| Add auxiliary goal prediction reward | Multi-agent policy · More complex environments | policy optimization | Helped |
| Train across varied tasks and environment configurations | Communication policy · Compositional language formation | compositionality | Helped |
Closest papers by meaning
- Emergent Coordination in Multi-Agent Language Models (ICLR 2026 · similarity 0.88)
- A compressive-expressive communication framework for compositional representations (NeurIPS 2025 · similarity 0.87)
- Language Grounded Multi-agent Reinforcement Learning with Human-interpretable Communication (NeurIPS 2024 · similarity 0.87)
- Bridging semantics and pragmatics in information-theoretic emergent communication (NeurIPS 2024 · similarity 0.86)
- Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · similarity 0.86)
- CtD: Composition through Decomposition in Emergent Communication (ICLR 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.