Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Trading off Utility, Informativeness, and Complexity in Emergent Communication
Mycal Tucker, Roger P. Levy, Julie Shah, Noga Zaslavsky
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 |
|---|---|---|---|
| anneal complexity weight lambda_C to move through complexity range | VQ-VIB · color reference game | human-agent gNID: 0.18 (0.10) -> 0.16 (0.01) | Improved |
| train with utility only (lambda_I = lambda_C = 0) instead of IB pressures (ablation) | VQ-VIB · color reference game | human-agent gNID: 0.16 (0.01) -> 0.42 (0.02) | Hurt |
| train with utility only (ablation) | VQ-VIB · color reference game | efficiency loss: 0.03 (0.00) -> 0.11 (0.01) | Hurt |
| replace one-hot with VQ-VIB embedding | one-hot agents · color reference game | efficiency loss: 0.05 (0.00) -> 0.03 (0.00) | Improved |
| replace one-hot with VQ-VIB embedding | one-hot agents · color reference game | human-agent gNID: 0.16 (0.00) -> 0.16 (0.01) | No clear effect |
| replace one-hot/Proto with VQ-VIB | Proto and one-hot · uniform environment | utility | Improved |
| increase codebook size K from 9 to 100 (ablation) | VQ-VIB · uniform environment | utility | Improved |
| increase codebook size K from 9 to 100 (ablation) | Proto and one-hot · uniform environment | utility | Improved |
| increase informativeness weight lambda_I with lambda_C = 0 (ablation) | VQ-VIB · uniform environment | utility | Improved |
| use REINFORCE instead of straight-through estimator (ablation) | VQ-VIB · color reference game | utility | Hurt |
Closest papers by meaning
- Emergent Discrete Communication in Semantic Spaces (NeurIPS 2021 · similarity 0.89)
- Bridging semantics and pragmatics in information-theoretic emergent communication (NeurIPS 2024 · similarity 0.89)
- A compressive-expressive communication framework for compositional representations (NeurIPS 2025 · similarity 0.89)
- Lewis's Signaling Game as beta-VAE For Natural Word Lengths and Segments (ICLR 2024 · similarity 0.88)
- Semantics and Spatiality of Emergent Communication (NeurIPS 2024 · similarity 0.87)
- Unsupervised Translation of Emergent Communication (AAAI 2025 · similarity 0.87)
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.
