LGAINov 4, 2025

Test-time Adaptation of Tiny Recursive Models

arXiv:2511.02886v14 citationsh-index: 21Has Code
Originality Synthesis-oriented
AI Analysis

This work addresses compute constraints for participants in the ARC Prize competition, but it is incremental as it builds on existing tiny recursive models with a focus on efficient adaptation.

The paper tackled the problem of adapting tiny recursive models to competition tasks within strict compute limits by pre-training on public ARC tasks and then fine-tuning, achieving a score of 6.67% on semi-private evaluation tasks.

Prior to the close of the 2025 ARC Prize competition, the leading open source approach - known as TRM, or Tiny Recursive Models - involved training a 7M parameter recursive neural network on augmented variants of ARC tasks. That approach scored approximately 7.8% on the public ARC AGI II evaluation set, but required a level of compute far in excess of what is allowed during the competition. This paper shows that, by starting from a tiny recursive model that has been pre-trained on public ARC tasks, one can efficiently fine-tune on competition tasks within the allowed compute limits. Specifically, a model was pre-trained on 1,280 public tasks for 700k+ optimizer steps over 48 hours on 4xH100 SXM GPUs to obtain a ~10% score on the public evaluation set. That model was then post-trained in just 12,500 gradient steps during the competition to reach a score of 6.67% on semi-private evaluation tasks. Notably, such post-training performance is achieved by full-fine tuning of the tiny model, not LoRA fine-tuning or fine-tuning of task embeddings alone.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes