RCTS boosts vision-language question answering by adding generated reasoning contexts to a knowledge base and re-ranking retrieved examples with a Monte Carlo tree search.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Re-ranking Reasoning Context with Tree Search Makes Large Vision-Language Models Stronger
RCTS boosts vision-language question answering by adding generated reasoning contexts to a knowledge base and re-ranking retrieved examples with a Monte Carlo tree search.