ReCCur: A Recursive Corner-Case Curation Framework for Robust Vision-Language Understanding in Open and Edge Scenarios

2026-01-06

arXiv.org

AI Summary

This work addresses the failure of vision-language understanding in open and edge scenarios caused by rare or extreme corner cases. To this end, the authors propose a computationally efficient recursive multi-agent framework that integrates vision-language models with multiple encoders (CLIP, DINOv2, and BEiT). The approach enables automatic purification of noisy data and fine-grained, auditable annotation through tri-modal consistency filtering, region-to-semantic chain verification, mixture-of-experts knowledge distillation, and region-wise adversarial evidence labeling. With minimal human intervention, the system substantially enhances data purity and class separability while remaining deployable on consumer-grade GPUs, thereby providing a high-accuracy, interpretable foundation for downstream tasks.

Technology Category

• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Multimodal Learning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Corner cases are rare or extreme scenarios that drive real-world failures, but they are difficult to curate at scale: web data are noisy, labels are brittle, and edge deployments preclude large retraining. We present ReCCur (Recursive Corner-Case Curation), a low-compute framework that converts noisy web imagery into auditable fine-grained labels via a multi-agent recursive pipeline. First, large-scale data acquisition and filtering expands a domain vocabulary with a vision-language model (VLM), crawls the web, and enforces tri-modal (image, description, keyword) consistency with light human spot checks to yield refined candidates. Next, mixture-of-experts knowledge distillation uses complementary encoders (e.g., CLIP, DINOv2, BEiT) for kNN voting with dual-confidence activation and uncertainty sampling, converging to a high-precision set. Finally, region-evidence VLM adversarial labeling pairs a proposer (multi-granularity regions and semantic cues) with a validator (global and local chained consistency) to produce explainable labels and close the loop. On realistic corner-case scenarios (e.g., flooded-car inspection), ReCCur runs on consumer-grade GPUs, steadily improves purity and separability, and requires minimal human supervision, providing a practical substrate for downstream training and evaluation under resource constraints. Code and dataset will be released.

Problem

• corner cases
• vision-language understanding
• data curation
• edge scenarios
• open-world

Innovation

• corner-case curation
• vision-language model
• multi-agent recursive pipeline
• mixture-of-experts distillation
• adversarial labeling