Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space

TL;DR

LatentSeek introduces a test-time, latent space reasoning method for LLMs that improves performance through iterative policy gradient updates guided by self-generated rewards, without retraining the model.

Contribution

This paper presents LatentSeek, a novel framework leveraging latent space policy gradient for test-time reasoning, outperforming existing methods and demonstrating efficiency and scalability.

Findings

LatentSeek outperforms Chain-of-Thought prompting and fine-tuning on reasoning benchmarks.

It converges within a few iterations for average complexity problems.

The method is highly efficient and scalable for reasoning tasks.

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.