Beyond Parameter Finetuning: Test-Time Representation Refinement for Node Classification

TL;DR

This paper introduces TTReFT, a test-time representation fine-tuning framework for node classification that improves out-of-distribution performance by adapting latent representations rather than model parameters.

Contribution

It proposes a novel approach shifting adaptation from parameters to representations, with three key innovations and theoretical guarantees for OOD settings.

Findings

Consistently outperforms existing methods across five benchmarks.

Achieves superior OOD generalization in node classification tasks.

Provides theoretical guarantees for the proposed framework.

Abstract

Graph Neural Networks frequently exhibit significant performance degradation in the out-of-distribution test scenario. While test-time training (TTT) offers a promising solution, existing Parameter Finetuning (PaFT) paradigm suffer from catastrophic forgetting, hindering their real-world applicability. We propose TTReFT, a novel Test-Time Representation FineTuning framework that transitions the adaptation target from model parameters to latent representations. Specifically, TTReFT achieves this through three key innovations: (1) uncertainty-guided node selection for specific interventions, (2) low-rank representation interventions that preserve pre-trained knowledge, and (3) an intervention-aware masked autoencoder that dynamically adjust masking strategy to accommodate the node selection scheme. Theoretically, we establish guarantees for TTReFT in OOD settings. Empirically, extensive experiments across five benchmark datasets demonstrate that TTReFT achieves consistent and superior performance. Our work establishes representation finetuning as a new paradigm for graph TTT, offering both theoretical grounding and immediate practical utility for real-world deployment.