New research with @Tsinghua_Uni: Spatial-TTT.
A framework for streaming visual-based spatial intelligence with test-time training (TTT). Spatial-TTT adapts fast weights to capture and organize spatial evidence from long video streams, enabling models to build structured 3D spatial memory over time.
Highlights:
🔹Efficient streaming memory. Fast weights act as compact spatial memory with sublinear memory growth over 7000+ frames and more than 40% lower compute.
🔹Spatial-predictive mechanism. TTT layers with 3D spatiotemporal convolution capture geometric correspondence and temporal continuity.
🔹SOTA results on long-horizon video spatial understanding (VSI-Bench).
The paper ranked #1 on @huggingface Daily Papers on March 13.
Project page: https://t.co/Lw0WxIWgGr
GitHub: https://t.co/95c4EomcCa
Paper: https://t.co/TMW8LciRMn
Model & Data:
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
streaming — Sending a model's response token by token as it is generated, so the reader sees text immediately instead of waiting for the whole answer.
Why it matters
A concrete answer to unbounded context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → cost in streamingSending a model's response token by token as it is generated, so the reader sees text immediately instead of waiting for the whole answer.Full definition → video: adapt fast weights instead of growing a buffer, with numbers on both memory growth and long-horizon spatial accuracy.
