TTT3R: Test-Time Training Enhances Length Generalization in 3D Reconstruction
rsasaki0109 · x · 2026-07-31
This is an ICLR 2026 paper focusing on 3D reconstruction. While modern Recurrent Neural Networks (RNNs) are competitive due to linear-time complexity, their performance degrades significantly beyond the training context length.
The authors revisit 3D reconstruction foundation models from a Test-Time Training perspective, framing them as online learning problems. By leveraging alignment confidence between the memory state and incoming observations, they derive a closed-form learning rate for memory updates. This training-free intervention, termed TTT3R, effectively balances retaining historical information and adapting to new observations, substantially improving length generalization.
Related event: TTT3R: Enhancing 3D Reconstruction via Test-Time Training(2 posts)→
More from Multimodal
- MiniMax H3 Video Model Lands on Magnific with Multi-Image and Motion Controls — aziz4ai · 2026-07-31
- MiniMax H3 Video Model Lands on Leonardo with Native Audio Generation — aziz4ai · 2026-07-31
- MiniMax H3 Video Model Hits Runware with Native Synced Audio — aziz4ai · 2026-07-31
- Single Image to Multi-Cam Video: Testing Seedream 5.0 Pro & Seedance 2.5 — aziz4ai · 2026-07-31
- ByteDance's Seedance 2.5 Targets Industrial Use: Autonomous Vehicle Simulation — hey_abusiddik · 2026-07-31
- Generating synced SFX for videos while preserving voiceover using SFX MCP — Soggy-Weather169 · 2026-07-31