Test-Time Meta-Adaptation with Self-Synthesis

Zeyneb N. Kaya, Nick Rui

Track: tiny / short paper (up to 5 pages)

Keywords: Meta-Learning, Synthetic Data, Test-Time Training

TL;DR: We propose MASS, a meta-learning framework that trains LLMs to self-adapt for a task at test time by generating targeted synthetic data and performing self-updates.

Abstract: As strong general reasoners, large language models (LLMs) encounter diverse domains and tasks, where the ability to adapt and self-improve at test time is valuable. We introduce MASS, a meta-learning framework that enables LLMs to self-adapt by generating problem-specific synthetic training data and performing targeted self-updates optimized for downstream performance at inference time. We train this behavior end-to-end via bilevel optimization: an inner loop adapts on self-generated examples while an outer loop meta-learns data-attribution signals and rewards post-update task performance. The synthetic data is optimized with scalable meta-gradients, backpropagating the downstream loss through the inner updates to reward useful generations. Experiments on mathematical reasoning show that MASS learns to synthesize per-instance curricula that yield effective, data-efficient test-time adaptation.

Anonymization: This submission has been anonymized for double-blind review via the removal of identifying information such as names, affiliations, and identifying URLs.

Submission Number: 31