Universal Test-Time Training

cs.LG, cs.CL, cs.CV

Submitted: 2026-10-04

Updated: 2026-10-04

Comments: 37 pages. Project page: https://zefan-cai.github.io/uTTT.github.io/ ; code: https://github.com/Zefan-Cai/uTTT

Code: https://github.com/Zefan-Cai/uTTT

Project page: https://zefan-cai.github.io/uTTT.github.io

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• Longformer: The Long-Document Transformer
• Mixture of Universal Experts: Scaling Virtual Width via Depth-Width Transformation
• Addressing Some Limitations of Transformers with Feedback Memory
• The Pile: An 800GB Dataset of Diverse Text for Language Modeling
• Path-Constrained Mixture-of-Experts
• UniPool: Learning Expert-to-Layer Ownership from Brief Global Access
• Tying the Loop -- Tied Expert Layers in Mixture-of-Experts Language Models
• Kimi Linear: An Expressive, Efficient Attention Architecture
• GLU Variants Improve Transformer
• Learning to (Learn at Test Time)
• ReXMoE: Reusing Experts with Minimal Overhead in Mixture-of-Experts
• End-to-End Test-Time Training for Long Context
• Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks