Assignment location influences how AI describes computation results
Editorial routing shapes how computational results are qualified in AI-assisted scientific writing
Computation and Language
Summary
Sometimes when AI systems write about computer-based science results, they leave out important details about how certain numbers compare. This happens depending on where the AI is told to keep those comparison details—in a group folder, extra document, or notes. The authors tested different AI versions and found that some styles keep these details better than others, but just reminding the AI to be accurate didn’t help. So, even if the information exists somewhere, it might not appear where readers expect it in the writing.
What this means in practice
- •For scientific writing teams: Design manuscript workflows to place comparison details where AI is more likely to report them alongside computational results.
- •For ai system developers: Improve AI-assisted writing tools by targeting routing rules that keep numerical qualifications in visible text rather than external repositories.
Authors
Jihan Kim
Abstract
Large language models increasingly analyze computational results and draft manuscripts, making reliable communication as important as correct analysis. Using fixed computational evidence, we tested whether assigning comparisons across modeling choices elsewhere in a research workflow changes manuscript reporting. In constrained sentence-writing tasks, Anthropic's Claude Sonnet 5 often omitted numerical qualifications when detailed comparisons were assigned to a group repository, but retained them more often when the same comparison was assigned to Supporting Information or its own working notes; Claude Opus 5 was less sensitive. These effects did not follow a simple accessibility ordering. A targeted placement rule largely restored sentence-level qualification, whereas a generic accuracy reminder did not. Longer contributions retained numerical qualifications, although some summaries across computational settings were still redirected to the repository. Thus, documenting context within an AI workflow does not ensure its communication where readers encounter the result.