Preprint
Effects of Prompt Length on Domain-specific Tasks for Large Language Models
Qibang Liu et al., arXiv preprint 2502.14255, 2025
Read the originalRead on October 17, 2026
- What it is
- An experiment that tests language models on domain-specific tasks such as financial sentiment and monetary policy with prompts of different lengths, from the original short prompts to versions that add background knowledge. It finds that longer prompts generally help and that very short prompts can hurt.
- How much weight it carries
- A preprint that has not been through peer review, from three university researchers. Its results come from a fixed set of classification tasks, so they show a trend rather than a rule for every chat request. It is one of the few controlled tests of prompt length, which is why it is cited with its limits.
Facts taken from this source
- The same 2025 study found that even with extensive background knowledge in the prompt, language models still struggled with the hardest domain-specific tasks.
“despite providing extensive background knowledge of the prompts, LLMs still struggle with challenging domain-specific tasks” (V. Conclusions)
- A 2025 study of language models on domain-specific tasks found that longer prompts generally improved performance, while shorter prompts could make it worse.
“Our findings indicate that longer prompts generally enhance model performance, while shorter prompts can be detrimental.” (V. Conclusions)