An inference-time technique that uses token activation dynamics to adaptively restrict text attention to important visual tokens, improving VLM accuracy on VQA, grounding, counting, OCR, and hallucination benchmarks.
ChartQA: A benchmark for question answering about charts with visual and logical reasoning
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
-
Aligning What Vision-Language Models See and Perceive with Adaptive Information Flow
An inference-time technique that uses token activation dynamics to adaptively restrict text attention to important visual tokens, improving VLM accuracy on VQA, grounding, counting, OCR, and hallucination benchmarks.