Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space
Hengli Li, Chenxi Li, Tong Wu, Xuekai Zhu, Yuxuan Wang, Zhaoxin Yu, Eric Hanchen Jiang, Song-Chun Zhu, Zixia Jia, Ying Nian Wu, Zilong Zheng
TL;DR
LatentSeek introduces test-time instance-level adaptation in latent space to boost reasoning without updating model parameters. It formulates a policy-gradient search over latent representations, guided by self-generated rewards, and demonstrates strong gains on GSM8K, MATH-500, and AIME2024 across diverse backbones. The approach achieves rapid convergence (often under a few iterations) and shows favorable generalization and efficiency compared to parameter-tuning or token-space methods. The work highlights latent-space TTIA as a practical, scalable path toward enhanced reasoning in LLMs and invites future refinements in reward modeling and larger-scale validation.
Abstract
Reasoning ability, a core component of human intelligence, continues to pose a significant challenge for Large Language Models (LLMs) in the pursuit of AGI. Although model performance has improved under the training scaling law, significant challenges remain, particularly with respect to training algorithms, such as catastrophic forgetting, and the limited availability of novel training data. As an alternative, test-time scaling enhances reasoning performance by increasing test-time computation without parameter updating. Unlike prior methods in this paradigm focused on token space, we propose leveraging latent space for more effective reasoning and better adherence to the test-time scaling law. We introduce LatentSeek, a novel framework that enhances LLM reasoning through Test-Time Instance-level Adaptation (TTIA) within the model's latent space. Specifically, LatentSeek leverages policy gradient to iteratively update latent representations, guided by self-generated reward signals. LatentSeek is evaluated on a range of reasoning benchmarks, including GSM8K, MATH-500, and AIME2024, across multiple LLM architectures. Results show that LatentSeek consistently outperforms strong baselines, such as Chain-of-Thought prompting and fine-tuning-based methods. Furthermore, our analysis demonstrates that LatentSeek is highly efficient, typically converging within a few iterations for problems of average complexity, while also benefiting from additional iterations, thereby highlighting the potential of test-time scaling in the latent space. These findings position LatentSeek as a lightweight, scalable, and effective solution for enhancing the reasoning capabilities of LLMs.