LLMind uses bio-inspired non-uniform sampling via a Mobius module and closed-loop semantic feedback to retain 82-97% of full-resolution VLM performance with only 1-5% of pixels on VQA benchmarks.
The geometry of m ¨obius transformations
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
-
LLMind: Bio-inspired Training-free Adaptive Visual Representations for Vision-Language Models
LLMind uses bio-inspired non-uniform sampling via a Mobius module and closed-loop semantic feedback to retain 82-97% of full-resolution VLM performance with only 1-5% of pixels on VQA benchmarks.