# On the Role of Temperature Sampling in Test-Time Scaling

arXiv: https://arxiv.org/abs/2510.02611 · HTML: https://mlsanity.com/paper/2510.02611/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 0 of its 102 references that are in the map.
- First posted on arXiv 2 Oct 2025.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Built on by (most cited first, 11 of 11)

- [What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity](https://mlsanity.com/paper/2511.15593/) (2025)
- [Distributional Biases in Post-Training: A Markovian Analysis of Reasoning Trajectories](https://mlsanity.com/paper/2511.07368/) (2025)
- [AgenticCache: Cache-Driven Asynchronous Planning for Embodied AI Agents](https://mlsanity.com/paper/2604.24039/) (2026)
- [Making Expert Reasoning Learnable with Self-Distillation](https://mlsanity.com/paper/2602.02405/) (2026)
- [Adaptive Test-Time Compute Allocation with Evolving In-Context Demonstrations](https://mlsanity.com/paper/2604.21018/) (2026)
- [Internalize the Temperature: On-Policy Self-Distillation as Policy Reheater for Reinforcement Learning](https://mlsanity.com/paper/2606.00755/) (2026)
- [Fork-Think with Confidence](https://mlsanity.com/paper/2606.31484/) (2026)
- [Temperature-Dependent Performance of Prompting Strategies in Extended Reasoning Large Language Models](https://mlsanity.com/paper/2604.08563/) (2026)
- [ReSET: Accurate Latency-Critical NVFP4 Reasoning via Step-Aware Temperature Scaling](https://mlsanity.com/paper/2606.13233/) (2026)
- [When Does the Best Sampling Temperature Rise with the Budget? Sufficient Conditions for Pass@k](https://mlsanity.com/paper/2608.14665/) (2026)
- [On the Convergence of Success Conditioning for Policy Optimization](https://mlsanity.com/paper/2610.03642/) (2026)

## Abstract (arXiv)

Large language models (LLMs) can improve reasoning at inference time through test-time scaling (TTS), where multiple reasoning traces are generated and the best one is selected. Prior work shows that increasing the number of samples K steadily improves accuracy. In this paper, we demonstrate that this trend does not hold indefinitely: at large K, further scaling yields no gains, and certain hard questions remain unsolved regardless of the number of traces. Interestingly, we find that different sampling temperatures solve different subsets of problems, implying that single-temperature scaling explores only part of a model's potential. We therefore propose scaling along the temperature dimension, which enlarges the reasoning boundary of LLMs. Averaged over Qwen3 (0.6B, 1.7B, 4B, 8B) and five representative reasoning benchmarks (AIME 2024/2025, MATH500, LiveCodeBench, Hi-ToM), temperature scaling yields an additional 7.3 points over single-temperature TTS. Temperature scaling also enables base models to reach performance comparable to reinforcement learning (RL)-trained counterparts, without additional post-training. We further provide a comprehensive analysis of this phenomenon and design a multi-temperature voting method that reduces the overhead of temperature scaling. Overall, our findings suggest that TTS is more powerful than previously thought, and that temperature scaling offers a simple and effective way to unlock the latent potential of base models.