Efficiently Scaling LLM Reasoning Programs with Certaindex

Test-time reasoning algorithms such as chain-of-thought, self-consistency, and MCTS enhance LLM problem-solving but can wastefully generate many tokens without improving accuracy. At the same time, we observe that these algorithms exhibit answer stabilization: their intermediate solutions often cease to change after a certain point, and further investment of compute does not change their final answer. To quantify this phenomenon, we introduce Certaindex, an algorithm-agnostic metric measuring this evolving stability, signaling when further computation is unlikely to alter the final result. Certaindex is lightweight, can accelerate reasoning program inference via early exit, and further enables dynamic token allocation, gang scheduling, and many opportunities when integrated with real-world LLM serving systems. To quantify real-world benefits, we built Certaindex as a scheduler into Dynasor, our reasoning-aware LLM serving system, and demonstrate up to 50% compute savings and 3.3× higher throughput in real workloads with no accuracy drop. Our code is available at https://github.com/hao-ai-lab/ Dynasor.git.

• Work done while interning at UCSD 39th Conference on Neural Information Processing Systems (NeurIPS 2025). Recent work has shown that allocating more compute at inference time, so-called "test-time scaling", consistently boosts performance on hard reasoning tasks [8; 14; 15]. Methods such as Chain-of-Thought (CoT) [16], Best-of-N sampling [17; 18], Self-Consistency (SC) [4], and search-based