Abstract
Onboard vision-language models could enable satellites to answer queries directly, but exhaustive tiled inference over high-resolution imagery is slow and energy-intensive. We identify answer-invariant token redundancy (AITR): image tiles and vision tokens that can be removed without changing the final answer. We present Rift, a two-stage system that performs query-conditioned tile pruning followed by elastic prefill to reduce token budget. We evaluate it on LLaVA-1.5 7B running on Jetson AGX Orin. Compared with exhaustive tiled inference, Rift reduces energy by 78% and latency by 69%, while increasing accuracy from 45% to 73%.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Janveja, I., Zhang, D., Oh, S., & Vasisht, D. (2026). Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge. https://omanscience.com/en/articles/exploiting-answer-invariant-redundancies-in-satellite-imagery-for-efficient-vlm-inference-on-edge
MLA 9
Janveja, Ishani, et al. "Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge." https://omanscience.com/en/articles/exploiting-answer-invariant-redundancies-in-satellite-imagery-for-efficient-vlm-inference-on-edge.
Chicago (author–date)
Janveja, Ishani, Davis Zhang, Seoyul Oh, and Deepak Vasisht. 2026. "Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge." https://omanscience.com/en/articles/exploiting-answer-invariant-redundancies-in-satellite-imagery-for-efficient-vlm-inference-on-edge.
Harvard
Janveja, I., Zhang, D., Oh, S. and Vasisht, D. (2026) 'Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge', Available at: https://omanscience.com/en/articles/exploiting-answer-invariant-redundancies-in-satellite-imagery-for-efficient-vlm-inference-on-edge.
Vancouver
Janveja I, Zhang D, Oh S, Vasisht D. Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge. https://omanscience.com/en/articles/exploiting-answer-invariant-redundancies-in-satellite-imagery-for-efficient-vlm-inference-on-edge
IEEE
I. Janveja, D. Zhang, S. Oh, and D. Vasisht, "Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge," https://omanscience.com/en/articles/exploiting-answer-invariant-redundancies-in-satellite-imagery-for-efficient-vlm-inference-on-edge.