FunnelAL: Retrieve-then-Rank Active Learning for Single-Class Discovery
cs.CV, cs.IR, cs.LG
Submitted: 2026-07-28
Updated: 2026-10-27
Comments: 36 pages, 5 figures. v2 corrects an ablation implementation error and updates SEALS and AnchorAL to better align with their published algorithms. These changes affect some reported comparisons and narrow our conclusions about the benefit of exploration. Adds iNaturalist, fMoW and CLIP evaluations
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Revisiting Active Learning in the Era of Vision Foundation Models
- xView: Objects in Context in Overhead Imagery
- Eliciting Human Preferences with Language Models
- Large Language Models for Data Annotation and Synthesis: A Survey
- Positive-First Most Ambiguous: A Simple Active Learning Criterion for Interactive Retrieval of Rare Categories
- Fine-Grained Visual Classification of Aircraft
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs