Search papers, labs, and topics across Lattice.
This paper introduces Reflection Steering, a training-free framework that disentangles reflection-related activations from general reasoning in large language models (LLMs) to optimize token efficiency. By contrasting reflective and non-reflective hidden states, applying PCA for denoising, and orthogonalizing against reasoning directions, the method achieves a significant reduction in reasoning tokens used during inference. Experimental results demonstrate an average reduction of 16.9% in reasoning tokens across multiple benchmarks, while also allowing for adjustable intervention strengths to balance efficiency and accuracy.
Reflection Steering cuts reasoning token usage by nearly 17% while maintaining accuracy, revolutionizing how LLMs handle reflection during inference.
Large reasoning models often produce reasoning traces with verification, revision, and backtracking. When reflection merely re-checks established results, it wastes reasoning tokens and increases latency. Most existing reflection steering methods add a label-derived mean-difference direction across preset layers, but its entanglement with reasoning and length signals destabilizes the accuracy-efficiency trade-off. In this paper, we propose Reflection Steering, a training-free framework for controlling reflection-associated computation within LLMs by disentangling reflection-related activations from general reasoning. Specifically, we contrast reflective and non-reflective hidden states at each LLM layer, denoise the resulting reflection directions with PCA, and orthogonalize them against general-reasoning directions. To limit downstream amplification from early-layer interventions, we calibrate each layer across multiple intervention strengths on a small set, retain only stable layers, and apply bounded projection removal to their residual-stream activations. We conduct extensive experiments across two public benchmarks and three open-weight LLMs against state-of-the-art activation-steering baselines. Results show that Reflection Steering reduces reasoning tokens by 16.9% on average across six matched settings. Besides, our method further introduces a bounded reflection intervention-strength parameter $\alpha$, enabling deployment-time adjustment to balance token savings, accuracy, and generation stability.
