Research map / Reinforcement learning / Visual and representation RL
Learning Actionable Information Landscapes for Multimodal Active Sensing in Hawkmoths
Abdo Sharafeldin, Yaqing Wang, Simon Sponberg, Hannah Choi
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a multimodal predictive-coding approach that learns action-conditioned information landscapes for active sensing, modeled on hawkmoth behavior.
Problem
How animals' action-conditioned information maps for guiding exploration are learned from multimodal sensory experience, given that prior studies were confined to unimodal sensing.
Approach
A multimodal predictive-coding framework that learns action-conditioned information landscapes from multisensory experience, extending prior unimodal active-sensing models to multiple modalities.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add uncertainty-gated lexicographic switching versus exploit-only, epsilon-greedy, fixed-time and non-lexicographic controls | nectary-search agent · nectary search (funnel and flat+cross flowers) | training success rate | Improved |
| adaptive gating versus non-lexicographic control on unpatterned flat flowers | nectary-search agent · nectary search, flat flower without pattern | final success rate | No clear effect |
| adaptive switching versus fixed-time switching on funnel flowers under curvature shift | nectary-search agent · curvature shift validation, funnel flowers | mean success | No clear effect |
| uncertainty gating under visual degradation on funnel flowers | nectary-search agent · sensory degradation validation, funnel flowers | mean success | Hurt |
| adaptive switching under all three shifts on flat flowers | nectary-search agent · shift validation, flat flowers | mean success | Improved |
| use weighted-unimodal uncertainty instead of aggregate entropy (ablation) | uncertainty-gated controller · flat flowers with cross | success | Hurt |
| add a cross pattern to flat flowers (ablation) | nectary-search agent · flat flowers | performance | Improved |
| compare biological probing trajectories to matched random-heading trajectories | learned UR maps · Deora et al. probing data | realized weighted UR | Improved |
| real moth trajectories versus shuffled and exact-flower controls | flower-tracking UR maps · Stockl et al. tracking data | time-averaged realized UR | Improved |
Closest papers by meaning
- Through Hawks’ Eyes: Synthetically Reconstructing the Visual Field of a Bird in Flight (IJCV 2023 · similarity 0.85)
- Theory of Space: Can Foundation Models Construct Spatial Beliefs through Active Exploration? (ICLR 2026 · similarity 0.85)
- MABe22: A Multi-Species Multi-Task Benchmark for Learned Representations of Behavior (ICML 2023 · similarity 0.84)
- Intrinsic Goals for Autonomous Agents: Model-Based Exploration in Virtual Zebrafish Predicts Ethological Behavior and Whole-Brain Dynamics (NeurIPS 2025 · similarity 0.83)
- BioCLIP 2: Emergent Properties from Scaling Hierarchical Contrastive Learning (NeurIPS 2025 · similarity 0.83)
- Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.83)
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.
