research
Fine Detail Leaves the Global Attention Queue
SpecSlot improved ImageNet-512 FID by 2.84 over a compute-matched control with 3.9 percent fewer FLOPs.
Summary
SpecSlot improved ImageNet-512 FID by 2.84 over a compute-matched control with 3.9 percent fewer FLOPs.
Pixel-space image generators usually buy finer detail with more globally attended tokens. SpecSlot separates spatial anchors into frequency-band slots and sends only the lowest-frequency slot from each anchor through global attention; higher-frequency detail stays local. On ImageNet-512, the paper reports a 2.84 FID improvement over a FLOPs-matched control with 3.9 percent fewer FLOPs, and a result within 0.69 FID of JiT-B/16 at 38.2 percent lower cost. Text-to-image tests also improved DPG-Bench by 1.06 over JiT-B/8 at 40.6 percent lower forward cost. Those comparisons cover the reported models and benchmarks, not every generation workload.
Why it matters
SpecSlot improved ImageNet-512 FID by 2.84 over a compute-matched control with 3.9 percent fewer FLOPs.
Limits and context
- SpecSlot separates spatial anchors into frequency-band slots and sends only the lowest-frequency slot from each anchor through global attention; higher-frequency detail stays local.
- Those comparisons cover the reported models and benchmarks, not every generation workload.
Key claims
SpecSlot improved ImageNet-512 FID by 2.84 over a compute-matched control with 3.9 percent fewer FLOPs.
Qualification: SpecSlot separates spatial anchors into frequency-band slots and sends only the lowest-frequency slot from each anchor through global attention; higher-frequency detail stays local.
Evidence: source-2026-11-01-004
Sources
- arXiv preprint 2610.40647, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.