Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Learning to Communicate and Solve Visual Blocks-World Tasks
Qi Zhang, Richard Lewis, Satinder Singh, Edmund Durfee
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 |
|---|---|---|---|
| replace random-string utterances with utterances of frequent N-grams (ablation) | RL speaker-listener · grammatical actions, B=15 | mean grammatical actions: 3.15 -> 7.60 | Improved |
| replace random-string utterances with utterances of frequent N-grams (ablation) | RL speaker-listener · grammatical actions, B=25 | mean grammatical actions: 2.75 -> 7.14 | Improved |
| replace random-string utterances with utterances of frequent N-grams (ablation) | Bandit speaker-listener · grammatical actions, B=15 | mean grammatical actions: 4.33 -> 10.11 | Improved |
| replace random-string utterances with utterances of frequent N-grams (ablation) | Bandit speaker-listener · grammatical actions, B=25 | mean grammatical actions: 5.47 -> 7.24 | Improved |
| replace random-string utterances with utterances of frequent N-grams (ablation) | SL speaker-listener · grammatical actions, B=15 | mean grammatical actions: 6.29 -> 8.06 | Improved |
| replace random-string utterances with utterances of frequent N-grams (ablation) | SL speaker-listener · grammatical actions, B=25 | mean grammatical actions: 6.34 -> 6.04 | No clear effect |
| add entropy regularization to SL training (ablation) | SL speaker-listener · extrapolation, partial completion | partial completion | Improved |
| train with Bandit instead of SL | speaker-listener · extrapolation | partial completion | Improved |
| train with Bandit or RL instead of SL | single-agent baseline · interpolation/extrapolation | full and partial completion | Improved |
| use speaker-listener instead of single agent | blocks-world agents · extrapolation | full and partial completion | Hurt |
| sample listener actions from softmax instead of greedy | Bandit listener · B=25 partial completion | partial completion | No clear effect |
| sample listener actions from softmax instead of greedy | SL listener · B=25 partial completion | partial completion | No clear effect |
Closest papers by meaning
- Learning Multi-Object Positional Relationships via Emergent Communication (AAAI 2024 · similarity 0.89)
- A compressive-expressive communication framework for compositional representations (NeurIPS 2025 · similarity 0.88)
- Bridging semantics and pragmatics in information-theoretic emergent communication (NeurIPS 2024 · similarity 0.87)
- Iterated Learning Improves Compositionality in Large Vision-Language Models (CVPR 2024 · similarity 0.87)
- Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · similarity 0.87)
- RLZero: Direct Policy Inference from Language Without In-Domain Supervision (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.
