Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Modeling Conceptual Understanding in Image Reference Games
Rodolfo Corona Rodriguez, Stephan Alaniz, Zeynep Akata
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 |
|---|---|---|---|
| use Epsilon Greedy attribute selection during practice instead of the Reactive policy | speaker · AwA2, SUN Attribute, and CUB test sets | average reward | Improved |
| use Active attribute selection during practice instead of the Reactive policy | speaker · AwA2, SUN Attribute, and CUB test sets | average reward | Improved |
| direct practice-game queries toward identifying misunderstood attributes | Epsilon Greedy and Active speakers · image reference game; first 20 practice games | practice games needed to identify misunderstood attributes: at least five times faster | Improved |
| add an agent embedding to the attribute-selection input (ablation) | Epsilon Greedy speaker · AwA2, SUN Attribute, and CUB; ALE and PNAS perception modules | average reward | Improved |
| use Epsilon Greedy or Active selection rather than Random Sampling during practice | speaker agent embeddings · listener-cluster recovery | variation of information | Improved |
| use a learned policy rather than random cluster assignment | speaker agent embeddings · listener-cluster recovery | variation of information | Improved |
| use Active selection instead of Epsilon Greedy selection | speaker · listener-cluster recovery | variation of information | No clear effect |
| use different perceptual modules for speaker and listeners | Epsilon Greedy speaker (ALE) and listener population (PNAS) · image reference game, including AwA2 | average reward | Hurt |
| use Epsilon Greedy selection rather than random attribute selection | speaker with ALE perception and listeners with PNAS perception · image reference game | average reward | Improved |
Closest papers by meaning
- Diagnosing the Compositional Knowledge of Vision Language Models from a Game-Theoretic View (ICML 2024 · similarity 0.87)
- GAM-Agent: Game-Theoretic and Uncertainty-Aware Collaboration for Complex Visual Reasoning (NeurIPS 2025 · similarity 0.86)
- What if Agents Could Imagine? Reinforcing Open-Vocabulary HOI Comprehension through Generation (NeurIPS 2026 · similarity 0.86)
- VS-Bench: Evaluating VLMs for Strategic Abilities in Multi-Agent Environments (CVPR 2026 · similarity 0.86)
- Learning Visually Grounded Domain Ontologies via Embodied Conversation and Explanation (AAAI 2025 · similarity 0.86)
- Learning Multi-Object Positional Relationships via Emergent Communication (AAAI 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.
