PeRL applies reinforcement learning to a vision-language model with image-order permutation and difficulty-based data filtering, improving multi-image reasoning while keeping single-image performance.
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
-
PeRL: Permutation-Enhanced Reinforcement Learning for Interleaved Vision-Language Reasoning
PeRL applies reinforcement learning to a vision-language model with image-order permutation and difficulty-based data filtering, improving multi-image reasoning while keeping single-image performance.