Slim-SC: Thought Pruning for Efficient Scaling with Self-Consistency
This work addresses the high computational cost of Self-Consistency for LLM reasoning, offering an efficient alternative for deployment, though it is incremental as it builds on existing TTS techniques.
The paper tackles the computational inefficiency of Self-Consistency in test-time scaling for LLMs by proposing Slim-SC, a pruning strategy that reduces inference latency by up to 45% and KVC usage by up to 26% while maintaining or improving accuracy on STEM reasoning datasets.
Recently, Test-Time Scaling (TTS) has gained increasing attention for improving LLM reasoning performance at test time without retraining the model. A notable TTS technique is Self-Consistency (SC), which generates multiple reasoning chains in parallel and selects the final answer via majority voting. While effective, the order-of-magnitude computational overhead limits its broad deployment. Prior attempts to accelerate SC mainly rely on model-based confidence scores or heuristics with limited empirical support. For the first time, we theoretically and empirically analyze the inefficiencies of SC and reveal actionable opportunities for improvement. Building on these insights, we propose Slim-SC, a step-wise pruning strategy that identifies and removes redundant chains using inter-chain similarity at the thought level. Experiments on three STEM reasoning datasets and two recent LLM architectures show that Slim-SC reduces inference latency and KVC usage by up to 45% and 26%, respectively, with R1-Distill, while maintaining or improving accuracy, thus offering a simple yet efficient TTS alternative for SC.