Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Catalytic Role Of Noise And Necessity Of Inductive Biases In The Emergence Of Compositional 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 |
|---|---|---|---|
| add noisy channel with noise 0.1 | signaling game agents · shapes3d (2 features, message length 2) | topographic similarity: 0.79 -> 0.87 | Helped |
| increase noise beyond the optimum (ablation) | signaling game agents · shapes3d (2 features, message length 2) | accuracy and topographic similarity | Hurt |
| use alphabet of 8 symbols with small noise (up to 0.08) (ablation) | signaling game agents · shapes3d (2 features, message length 2) | topographic similarity: 0.79 baseline | No clear effect |
| use alphabet of 8 symbols with noise 0.24 (ablation) | signaling game agents · shapes3d (2 features, message length 2) | topographic similarity: 0.79 -> 0.88 | Helped |
| restrict alphabet to 4 symbols (16 messages) (ablation) | signaling game agents · shapes3d (2 features, message length 2) | topographic similarity | Hurt |
| increase noise from 0 to 0.15 later in training (T=200 vs 2700) (ablation) | signaling game agents · shapes3d (2 features, message length 2) | topographic similarity: 0.85 -> 0.9 | Helped |
| bigger CNN in sender (16 filters) (ablation) | signaling game agents · shapes3d (2 features, message length 2) | topographic similarity: 0.86 at noise 0.14 vs 0.79 no noise | No clear effect |
| scramble input images into finer tiles (ablation) | signaling game agents · shapes3d (2 features, message length 2) | accuracy: 0.97 -> 0.53 (x=8) | Hurt |
| permute color-shape labels in factorized loss (ablation) | signaling game agents · shapes3d (2 features, message length 2) | compositionality w.r.t. standard features | Hurt |
| add noise to channel (ablation) | agents with permutation-based noise · shapes3d (2 features, message length 2) | topographic similarity | Mixed |
Closest papers by meaning
- Lewis's Signaling Game as beta-VAE For Natural Word Lengths and Segments (ICLR 2024 · similarity 0.87)
- A compressive-expressive communication framework for compositional representations (NeurIPS 2025 · similarity 0.87)
- Semantics and Spatiality of Emergent Communication (NeurIPS 2024 · similarity 0.87)
- CtD: Composition through Decomposition in Emergent Communication (ICLR 2025 · similarity 0.85)
- Speaking Your Language: Spatial Relationships in Interpretable Emergent Communication (NeurIPS 2024 · similarity 0.85)
- Towards a Formal Theory of Representational Compositionality (ICML 2025 · similarity 0.84)
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.