VisionFoundry generates synthetic VQA data from task keywords alone using LLMs and T2I models, delivering +7% on MMVP and +10% on CV-Bench-3D for VLMs.
Title resolution pending
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
-
VisionFoundry: Teaching VLMs Visual Perception with Synthetic Images
VisionFoundry generates synthetic VQA data from task keywords alone using LLMs and T2I models, delivering +7% on MMVP and +10% on CV-Bench-3D for VLMs.