Research map / Multi-agent systems and games / Cooperative multi-agent RL
Biases for Emergent Communication in Multi-agent Reinforcement Learning
Tom Eccles, Yoram Bachrach, Guy Lever, Angeliki Lazaridou, Thore Graepel
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 positive signalling loss | Independent REINFORCE agents · Summing MNIST digits | proportion of good runs: 0.00 -> 0.98 | Improved |
| Add positive listening loss | Independent REINFORCE agents · Summing MNIST digits | proportion of good runs: 0.00 -> 0.88 | Improved |
| Add both positive signalling and positive listening losses | Independent REINFORCE agents · Summing MNIST digits | proportion of good runs: 0.00 -> 1.00 | Improved |
| Add speaker-side social influence reward | Independent REINFORCE agents · Summing MNIST digits | proportion of good runs: 0.00 -> 0.00 | No clear effect |
| Add positive signalling loss | Independent A3C agents with V-trace · Treasure Hunt | proportion good: 0.28 -> 0.84 | Improved |
| Add positive listening loss | Independent A3C agents with V-trace · Treasure Hunt | proportion good: 0.28 -> 0.64 | Improved |
| Add both positive signalling and positive listening losses | Independent A3C agents with V-trace · Treasure Hunt | final reward: 12.45 \pm 0.48 -> 15.14 \pm 0.33 | Improved |
| Add speaker-side social influence reward | Treasure Hunt agents · Treasure Hunt | task reward | No clear effect |
| Use single-timestep CIC instead of multiple-timestep CIC (ablation) | Positive listening · Treasure Hunt | task reward | No clear effect |
| Set the cross-entropy loss weight to zero (ablation) | Positive listening · Summing MNIST digits | final reward | Hurt |
| Set the cross-entropy loss weight to zero (ablation) | Positive listening · Treasure Hunt | task reward | No clear effect |
Closest papers by meaning
- Emergent Communication through Negotiation (ICLR 2018 · similarity 0.90)
- Emergent Communication: Generalization and Overfitting in Lewis Games (NeurIPS 2022 · similarity 0.89)
- Multi-Agent Reinforcement Learning with Communication-Constrained Priors (NeurIPS 2025 · similarity 0.88)
- Learning Multi-Agent Communication with Contrastive Learning (ICLR 2024 · similarity 0.87)
- Lewis's Signaling Game as beta-VAE For Natural Word Lengths and Segments (ICLR 2024 · similarity 0.86)
- Semantics and Spatiality of Emergent Communication (NeurIPS 2024 · 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.
