Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Ease-of-Teaching and Language Structure from Emergent Communication
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 |
|---|---|---|---|
| Periodically replace the listener after 6k iterations | Referential game training · Referential game, 300k iterations | topographic similarity: 0.51 -> 0.59 | Helped |
| Periodically replace the listener after 6k iterations | Referential game training · Emergent languages | ease-of-teaching | Helped |
| Use a compositional mapping instead of a random permutation | Artificial language · Referential game | ease-of-teaching | Helped |
| Replace staggered listener resets with simultaneous resets (ablation) | Population training, N=2 · Referential game | ease-of-teaching | Mixed |
| Replace staggered listener resets with simultaneous resets (ablation) | Population training, N=2 · Referential game | topographic similarity: 0.53 -> 0.59 | Helped |
| Replace staggered listener resets with simultaneous resets (ablation) | Population training, N=10 · Referential game | ease-of-teaching | Helped |
| Replace staggered listener resets with simultaneous resets (ablation) | Population training, N=10 · Referential game | topographic similarity | Helped |
| Increase staggered listener population from N=2 to N=10 (ablation) | Staggered reset training · Referential game | topographic similarity: 0.57 -> 0.56 | Hurt |
| Increase staggered listener population from N=2 to N=10 (ablation) | Staggered reset training · Emergent languages | ease-of-teaching | Hurt |
| Use a larger listener population without resetting listeners (ablation) | No-reset training · Referential game | ease-of-teaching | Helped |
| Set speaker entropy coefficient to 0.05 instead of 0.1 (ablation) | Listener reset training · Referential game | ease-of-teaching | Hurt |
| Set listener entropy coefficient to 0.1 instead of 0.05 (ablation) | No-reset training · Referential game | training stability | Hurt |
Closest papers by meaning
- Emergence of Linguistic Communication from Referential Games with Symbolic and Pixel Input (ICLR 2018 · similarity 0.87)
- Emergence of Grounded Compositional Language in Multi-Agent Populations (AAAI 2018 · similarity 0.85)
- Emergence of Language with Multi-agent Games: Learning to Communicate with Sequences of Symbols (NeurIPS 2017 · similarity 0.85)
- Emergent Communication at Scale (ICLR 2022 · similarity 0.85)
- Speaking Your Language: Spatial Relationships in Interpretable Emergent Communication (NeurIPS 2024 · similarity 0.85)
- Lewis's Signaling Game as beta-VAE For Natural Word Lengths and Segments (ICLR 2024 · similarity 0.85)
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.