UniDoc-RL uses hierarchical actions and dense multi-rewards in a unified RL setup to let LVLMs progressively refine visual evidence from coarse retrieval to fine-grained cropping, delivering up to 17.7% gains over prior RL methods on three benchmarks.
- Evaluate the completeness and relevance of information in each image, and **select the single image most helpful for answering the query**
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
UniDoc-RL: Coarse-to-Fine Visual RAG with Hierarchical Actions and Dense Rewards
UniDoc-RL uses hierarchical actions and dense multi-rewards in a unified RL setup to let LVLMs progressively refine visual evidence from coarse retrieval to fine-grained cropping, delivering up to 17.7% gains over prior RL methods on three benchmarks.