Test-Time-Learning

Abstract: The main idea in test-time training (TTT) (Sun et al. 2024) is that a model with fixed parameters produces the supervision for another network that is updated during test-time (or inference-time). This article first reviews the TTT paper. Then, we discuss the problem with TTT and how LaCT addresses them, resulting in a powerful attention alternative that balances efficiency and performance. Currently, “test-time training” is an overloaded term with multiple meanings. In this article, we use the term to refer to the test-time training paradigm proposed in Sun et al. 2024, which is a framework for recurrent architectures for sequence modeling. ...

In DeltaProduct (Siems et al., 2025), they propose to improve DeltaNet (Yang et al., 2025) by updating the online memory with $n_h$ KVs for each token, which can be seen as performing multiple steps of gradient descent per token. I will explain how this method is almost the same as multi-KV DeltaNet and reveal a potential flaw in the design of DeltaProduct. Introduction DeltaNet We use row-vector notation. ...

This blog post is part 1 of a series that describes my attempt in implementing the Test-Time Training (TTT) model proposed by Sun et al. (2024), and Titans, proposed by Behrouz et al., (2024). At the time of writing, these two are two strong recurrent language models, but they have not yet open-sourced their implementation (TTT has only open-sourced the Jax implementation). Introduction to Test-Time Training Briefly explained, Test-Time Training (TTT) is an RNN model whose hidden state is replaced with an online learner, whose parameters are updated updated through gradient descent during inference. The goal is that this online learner compress contextual information into its parameters. A TTT operator can be expressed as: ...