Research map / Multi-agent systems and games / Human-AI and goal-conditioned
BAR — A Reinforcement Learning Agent for Bounding-Box Automated Refinement
Morgane Ayle, Jimmy Tekli, Julia El-Zini, Boulos El-Asmar, Mariette Awad
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 |
|---|---|---|---|
| apply BAR-DRL correction to PASCAL aeroplane detector proposals | Hierarchical detector · PASCAL VOC aeroplane | AP@50 (%): 7.9 -> 23.5 | Improved |
| apply BAR-DRL to human-annotator-biased boxes | FA-Kit Middle · Exp. 3 | AP@70: 0.3 -> 93.1 | Improved |
| apply BAR-DRL to human-annotator-biased boxes | FA-Kit Left · Exp. 3 | AP@70: 100 -> 96.1 | Hurt |
| apply BAR-DRL to RetinaNet proposals | RetinaNet · Exp. 4 Stickers | AP@75: 10.8 -> 64.2 | Improved |
| apply BAR-DRL to template matching proposals | Template matching · Exp. 5 | AP@50: 25.01 -> 88.5 (FA-Kit Right); 5.06 -> 8.9 (Stickers) | Mixed |
| apply BAR-CB (online contextual bandit) instead of none | Exp. 1 distribution init · Exp. 1 FA-Kit Right | AP@50: 84.64 -> 81.0 | Hurt |
| BAR-CB instead of BAR-DRL | BAR · Exp. 3 FA-Kit Middle | AP@70: 93.1 -> 50.8 | Hurt |
| raise beta from 0.70 to 0.80 (ablation) | BAR-DRL · Exp. 1 FA-Kit Right, 50 images | AP@70: 39.1 -> 52.5 | Improved |
| raise beta from 0.70 to 0.75/0.80 (ablation) | BAR-DRL · Exp. 1 FA-Kit Middle, 50 images | AP@50: 85.6 -> 79.5 (beta 0.75 -> 0.80) | Hurt |
| increase training images from 50 to 150 at beta 0.80 (ablation) | BAR-DRL · Exp. 1 FA-Kit Right | AP@70: 52.5 -> 82.1 | Improved |
Closest papers by meaning
- DeepEyesV2: Toward Agentic Multimodal Model (ICLR 2026 · similarity 0.84)
- Learning to Select Visual Tools from Experience (CVPR 2026 · similarity 0.84)
- MENTOR: Mixture-of-Experts Network with Task-Oriented Perturbation for Visual Reinforcement Learning (ICML 2025 · similarity 0.84)
- Breaking the SFT Plateau: Multimodal Structured Reinforcement Learning for Chart-to-Code Generation (ICLR 2026 · similarity 0.83)
- DMC-VB: A Benchmark for Representation Learning for Control with Visual Distractors (NeurIPS 2024 · similarity 0.83)
- SiT: Symmetry-invariant Transformers for Generalisation in Reinforcement Learning (ICML 2024 · 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.
