JiT

Search everything

Mentions · 7d2
Mentions · 30d4
All mentions4
First seenOct 13, 2026
Last mentioned3d ago

Timeline

4 articles · newest first

SpecSlot raises pixel-space detail without longer global attention

On ImageNet-512, SpecSlot improves FID by 2.84 versus a FLOPs-matched JiT control and uses 3.9% fewer FLOPs.

arXiv cs.AI / cs.LG / cs.CLOct 29, 17:59 UTC · 3d agoAI research 2 sources1

SpecSlot cuts global attention cost in pixel-space transformers

Only the lowest-frequency slot per spatial anchor attends globally; higher-frequency detail stays local.

Hugging Face daily papersupdated Oct 29, 17:59 UTC · 3d agofirst Oct 28, 20:00 UTC · 4d agoAI research 2 sources1

LILIE language-informed loss improves pixel-space text-to-image training

JiT models trained with LILIE cut GPIC gFD from 66.84 to 53.01 versus a pixel-only baseline.

Hugging Face daily papersOct 15, 20:00 UTC · 17d agoAI research

Correction training reduces diffusion models' training-inference gap

A corrector refines intermediate predictions and can be distilled back into the base model.

Hugging Face daily papersOct 13, 20:00 UTC · 19d agoAI research1 · 1 read

Appears with

SpecSlotai model · 2
LILIEai model · 1
SiTai model · 1
DiTai model · 1

Entities are extracted by the model from each article. First seen Oct 13, 2026, last mentioned 3d ago. Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.