利用潛在推理擴展測試時間計算：一種遞歸深度方法

Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach

February 7, 2025

作者: Jonas Geiping, Sean McLeish, Neel Jain, John Kirchenbauer, Siddharth Singh, Brian R. Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein

cs.AI

摘要

我們研究了一種新穎的語言模型架構，能夠通過在潛在空間中隱式推理來擴展測試時的計算。我們的模型通過迭代遞歸塊來工作，因此在測試時可以展開到任意深度。這與通過生成更多標記來擴展計算的主流推理模型形成對比。與基於思維鏈的方法不同，我們的方法不需要任何專門的訓練數據，可以使用較小的上下文窗口工作，並且可以捕捉在詞語中不容易表示的推理類型。我們將一個概念驗證模型擴展到了35億個參數和8000億個標記。我們展示了結果模型可以在推理基準測試中提高其性能，有時甚至可以達到相當於50億個參數的計算負載。

English

We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters.