preprint · arXiv (Cornell University) · 2024

Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters

Snell, Charlie, Lee, Jaehoon, Kelvin Xu, Aviral Kumar

Cited by 2,068Open access

Abstract

Enabling LLMs to improve their outputs by using more test-time computation is a critical step towards building generally self-improving agents that can operate on open-ended natural language. In this paper, we study the scaling of inference-time computation in LLMs, with a focus on answering the question: if an LLM is allowed to use a fixed but non-trivial amount of inference-time compute, how much can it improve its performance on a challenging prompt? Answering this question has implications not only on the achievable performance of LLMs, but also on the future of LLM pretraining and how one should tradeoff inference-time and pre-training compute. Despite its importance, little research attempted to understand the scaling behaviors of various test-time inference methods. Moreover, current work largely provides negative results for a number of these strategies. In this work, we analyze two primary mechanisms to scale test-time computation: (1) searching against dense, process-based verifier reward models; and (2) updating the model's distribution over a response adaptively, given the prompt at test time. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt. This observation motivates applying a "compute-optimal" scaling strategy, which acts to most effectively allocate test-time compute adaptively per prompt. Using this compute-optimal strategy, we can improve the efficiency of test-time compute scaling by more than 4x compared to a best-of-N baseline. Additionally, in a FLOPs-matched evaluation, we find that on problems where a smaller base model attains somewhat non-trivial success rates, test-time compute can be used to outperform a 14x larger model.

Cite this paper

APA 7

Snell, C., Lee, J., Xu, K., & Kumar, A. (2024). Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2408.03314

MLA 9

Snell, Charlie, et al. “Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters.” arXiv (Cornell University), 2024, https://doi.org/10.48550/arxiv.2408.03314.

Chicago

Snell, Charlie, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2024. “Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters.” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2408.03314.

Harvard

Snell, C. et al. (2024) ‘Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters’, arXiv (Cornell University). Available at: https://doi.org/10.48550/arxiv.2408.03314.

BibTeX

@misc{snell2024scaling,
  title = {{Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters}},
  author = {Snell, Charlie and Lee, Jaehoon and Xu, Kelvin and Kumar, Aviral},
  journal = {arXiv (Cornell University)},
  year = {2024},
  doi = {10.48550/arxiv.2408.03314}
}

Made from the index record. Check names, capitalisation and page numbers against the paper. Need another style? AI for Research has over 10,000.