Model Whisper: Steering Vectors Unlock Large Language Models' Potential in Test-time
This addresses the problem of computationally expensive and risky test-time adaptation for LLM users, offering a lightweight plug-and-play solution.
The paper tackles the challenge of efficiently unlocking large language models' reasoning potential for specific tasks without tuning model parameters, by introducing Test-Time Steering Vectors (TTSV) that steer models to higher confidence, achieving up to a 45.88% relative performance gain on the MATH500 task.
It is a critical challenge to efficiently unlock the powerful reasoning potential of Large Language Models (LLMs) for specific tasks or new distributions. Existing test-time adaptation methods often require tuning model parameters, which is not only computationally expensive but also risks degrading the model's pre-existing abilities.To address this, we introduce a lightweight component, Test-Time Steering Vectors (TTSV), which is prepended to the input while keeping the LLM's parameters entirely frozen. By optimizing the TTSV on test data to minimize the model's output entropy, we steer the model towards an internal state of higher confidence, activating its inherent abilities most relevant to the current task. TTSV is both lightweight and highly efficient to optimize, making it a true plug-and-play enhancement. Extensive experiments validate our approach's effectiveness on both base models and reasoning-enhanced models. For instance, on the MATH500 task, TTSV achieves a 45.88% relative performance gain on the Qwen2.5-Math-7B model and a 16.22% relative gain on the Qwen3-4B model. Furthermore, our approach exhibits robust generalization, with its steering vectors proving highly transferable across diverse tasks.