ImageNet

Headlines · 2

1. 3d
   SpecSlot separates image detail from global-attention sequence length

   A 22-page arXiv paper under review at ICLR 2027 introduces SpecSlot, a pixel-space rectified-flow Transformer designed to increase spectral detail without enlarging the global-attention sequence. It expands each spatial anchor into slots assigned to disjoint bands of DCT coefficients. Only the lowest-frequency slot participates in global attention, while all slots interact locally; a spatial head adds a local correction to the spectral prediction. The authors report comparable final FID and faster convergence versus a FLOPs-matched JiT control on ImageNet-256. On ImageNet-512, they report a 2.84 FID improvement with 3.9% fewer FLOPs, while on DPG-Bench the model reportedly exceeds JiT-B/8 by 1.06 at 40.6% lower forward cost. These results are author-reported and have not been independently verified.

   Hugging Face · Papers · 🔥 0
2. Oct 25
   Fei-Fei Li Says Spatial Intelligence Is AI's Next Frontier

   In an interview with historian Stephen Kotkin recorded at World Labs' San Francisco headquarters, Fei-Fei Li argued that large language models are a waypoint rather than AI's endpoint and that spatial intelligence is the next major frontier. She also warned about deepfakes, unequal distribution of AI-driven productivity gains, and the need to include educators and public institutions in technological change. The interview was recorded before AMD announced its approximately $8.2 billion all-stock acquisition of World Labs.

   虎嗅 AI · 🔥 0

Experience and discussion from the community

Share my ImageNet experienceAsk about ImageNet