A training-free RAG system with LLM-generated pseudo-TOCs and hierarchical retrieval reaches 64.7% on MMLongBench-Doc, surpassing prior SOTA by 15 points.
Visual Content AnalysisFor standalone images meeting size thresholds (100x100 pixels), we apply content-specific prompts: Single Image Prompt: 1Describe this visual content
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Zero-Shot Document Understanding using Pseudo Table of Contents-Guided Retrieval-Augmented Generation
A training-free RAG system with LLM-generated pseudo-TOCs and hierarchical retrieval reaches 64.7% on MMLongBench-Doc, surpassing prior SOTA by 15 points.