Papers by Matthew Renze
The Effect of Sampling Temperature on Problem Solving in Large Language Models (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Despite anecdotal reports, changes in temperature do not have a statistically significant impact on LLM performance for problem-solving tasks. |
| Approach: | They use a multiple-choice question-and-answer exam to investigate the effect of sampling temperature on LLM performance. |
| Outcome: | The results show that temperature changes do not have a statistically significant impact on LLM performance on problem-solving tasks. |