Research map / Reinforcement learning / Visual and representation RL
A Partially-Supervised Reinforcement Learning Framework for Visual Active Search
Anindya Sarkar, Nathan Jacobs, Yevgeniy Vorobeychik
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 |
|---|---|---|---|
| decompose the policy into prediction and search modules and update predictions from observed labels | VAS on xView · Helicopter target, C=25, single query | ANT: 0.54 -> 0.87 | Improved |
| meta-learn the prediction initialization jointly with the search policy | PSVAS on xView · Helicopter target, C=25, single query | ANT: 0.87 -> 0.92 | Improved |
| meta-learn the prediction initialization jointly with the search policy | PSVAS on DOTA · Ship target, C=25, single query | ANT: 2.46 -> 3.08 | Improved |
| learn interdependencies between sequential query choices instead of selecting the top-k grids greedily | MPS-VAS-topk · xView, Helicopter target, C=25, multi-query | ANT: 0.55 -> 0.63 | Improved |
| set supervised-loss weight λ to 0.1 rather than 0.001 (ablation) | PSVAS on xView · Small Car target, C=25 | ANT: 4.96 -> 5.51 | Improved |
| set supervised-loss weight λ to 0.1 rather than 0.001 (ablation) | MPS-VAS on xView · Small Car target, C=25 | ANT: 4.99 -> 5.55 | Improved |
| remove supervised prediction loss by setting λ=0 (ablation) | MPS-VAS on xView · Small Car target, C=25 | ANT: 5.55 -> 4.77 | Hurt |
| update prediction parameters from observed labels during search | VAS on DOTA · Ship target, C=25, single query | ANT: 2.46 -> 3.60 | Improved |
Closest papers by meaning
- Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search (ICLR 2026 · similarity 0.85)
- SenseSearch: Empowering Vision-Language Models with High-Resolution Agentic Search-Reasoning via Reinforcement Learning (CVPR 2026 · similarity 0.85)
- VRAG-RL: Empower Vision-Perception-Based RAG for Visually Rich Information Understanding via Iterative Reasoning with Reinforcement Learning (NeurIPS 2025 · similarity 0.85)
- DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization (ICLR 2024 · similarity 0.85)
- OpenSearch-VL: An Open Recipe for Frontier Multimodal Search Agents (NeurIPS 2026 · similarity 0.84)
- DeepSearch: Overcome the Bottleneck of Reinforcement Learning with Verifiable Rewards via Monte Carlo Tree Search (ICLR 2026 · 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.
