Fast ViT shows strong ImageNet results; scaling runs needed next
ducha_aiki · x · 2026-09-11
@duchaaiki shares at ECCV 2026 that Fast ViT works well on ImageNet, but scaling runs are still needed to earn @giffmana's blessing. A brief progress signal on the efficient vision architecture.
More from Research
- SlopCodeBench: Measuring how sloppy LLM-generated code really is — mitsuhiko · 2026-09-11
- Microsoft paper: read-only verification tools lift agent memory pass rate from 39% to 73% — dair_ai · 2026-09-11
- MIT Researcher Lands New Info Theory Result After 200+ Hours of Agent Work, 120 Pages to Review — blaizedsouza · 2026-09-11
- Yale NLP Releases IdeaAMBIG Benchmark Targeting Underspecified Research Ideas for LLMs — yale-nlp · 2026-09-11
- SAM-H uses SAM 2 masks for training-free homography tracking, +18.4pp on PlanarTrack — ducha_aiki · 2026-09-11
- Autoregressive vs Diffusion Training: The Data-Efficiency Math Behind Non-Causal Architectures — mgostIH · 2026-09-11