SciProdLLM 2025 is a forum for presenting and discussing research on integrating large language models (LLMs) into the typical research workflow: from ideation to experimentation to scientific writing, with a particular focus on human-centered approaches that ensure ethical and responsible use of LLMs. We also invite work that evaluates the quality of LLM-assisted research workflows and the resulting outputs.
Scope
We welcome submissions on any aspect of human–LLM collaboration for science production, evaluation of LLMs for science production, and/or evaluation of LLM-assisted scientific papers. Relevant topics include, but are not limited to, the following:
- Guiding idea generation through user feedback
- Automated experimentation following the experimental workflow used by human scientists (e.g., the workflow from data preprocessing to comparison to baselines)
- Human-curated datasets of scientific papers for fine-tuning LLMs for generating ideas and paper content (text, figures, tables, etc.) and publication ethics considerations for content generation
- LLM-assisted peer review and publication ethics considerations
- Benchmark datasets for evaluating LLMs on idea generation, experimentation, multimodal content generation, or scientific writing
- Evaluation metrics for detecting problematic papers (e.g., those containing suspicious citations or tortured phrases)
- Statistical analyses of collections of LLM-assisted papers (e.g., on topics, citations, or retractions)
- Ethics and safety of LLM in science production (e.g., authorship, bias, privacy and misuse)
Submission Instructions
SciProdLLM 2025 welcomes long and short papers. Long papers may consist of up to 8 pages of content, plus unlimited pages of references. Short papers may consist of up to 4 pages of content, plus unlimited pages of references. Both types of submissions must follow the same requirements and procedures as for IJCNLP–AACL 2025 main conference papers. Note that papers submitted as non-archival will be allocated presentation time at the workshop but will not be included in the proceedings. Submissions should also comply with the ARR publication ethics policy, where AI authorship is not permitted.
There are three supported submission modes:
- Direct submissions:
- Direct submissions will receive up to three double-blind reviews, and a final decision on acceptance from the workshop organizers. Direct submissions should be made through the SciProdLLM page on OpenReview.
- ARR submissions:
- Unpublished papers that have already been reviewed and meta-reviewed through ACL Rolling Review may be committed to SciProdLLM. These papers will not receive new reviews but may be meta-reviewed by the workshop organizers, who will make a final decision on acceptance. A commitment should be made through the SciProdLLM ARR Commitment page on OpenReview.
- Previously published papers:
- We invite non-archival submissions of papers that have already been recently published elsewhere. This allows such papers to gain more visibility from the workshop audience. To submit a previously published paper for presentation, please email SciProdLLM@groups.io with the details of your paper (title, authors, abstract, publication venue) and attaching a PDF copy of the paper. (Submissions of previously published papers need not adhere to the IJCNLP–AACL 2025 main conference paper policies on anonymity.)
Important Dates
All deadlines are 23:59 UTC−12 (“Anywhere on Earth”).
September 29 November 2, 2025: Submission deadline for direct submissions- October 27, 2025: ARR commitment deadline
November 3 November 5, 2025: Notification of acceptance of direct and ARR submissionsNovember 11 November 14, 2025: Camera-ready papers due- November 14, 2025: Non-archival submission deadline (for previously published papers)
- December 23, 2025: Workshop presentations
Contact Information
Email: SciProdLLM@groups.io