![[2602.20459] PreScience: A Benchmark for Forecasting Scientific Contributions](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20459] PreScience: A Benchmark for Forecasting Scientific Contributions
Summary
The paper introduces PreScience, a benchmark for forecasting scientific contributions using AI. It evaluates four generative tasks related to research prediction, utilizing a dataset of 98K AI-related papers.
Why It Matters
PreScience addresses the challenge of predicting future scientific advancements, which can enhance collaboration and research direction identification. This benchmark could significantly impact how researchers approach and prioritize their work, ultimately influencing the trajectory of scientific discovery.
Key Takeaways
- PreScience decomposes scientific forecasting into four tasks: collaborator prediction, prior work selection, contribution generation, and impact prediction.
- The dataset includes 98K AI-related papers with detailed metadata, enabling robust evaluation of forecasting models.
- Current AI models show moderate success in contribution generation, indicating room for improvement in simulating diverse and novel research.
- LACERScore, a new metric introduced, demonstrates improved measurement of contribution similarity compared to previous methods.
- The findings highlight the limitations of AI-generated research compared to human-authored work in terms of diversity and novelty.
Computer Science > Artificial Intelligence arXiv:2602.20459 (cs) [Submitted on 24 Feb 2026] Title:PreScience: A Benchmark for Forecasting Scientific Contributions Authors:Anirudh Ajith, Amanpreet Singh, Jay DeYoung, Nadav Kunievsky, Austin C. Kozlowski, Oyvind Tafjord, James Evans, Daniel S. Weld, Tom Hope, Doug Downey View a PDF of the paper titled PreScience: A Benchmark for Forecasting Scientific Contributions, by Anirudh Ajith and Amanpreet Singh and Jay DeYoung and Nadav Kunievsky and Austin C. Kozlowski and Oyvind Tafjord and James Evans and Daniel S. Weld and Tom Hope and Doug Downey View PDF Abstract:Can AI systems trained on the scientific record up to a fixed point in time forecast the scientific advances that follow? Such a capability could help researchers identify collaborators and impactful research directions, and anticipate which problems and methods will become central next. We introduce PreScience -- a scientific forecasting benchmark that decomposes the research process into four interdependent generative tasks: collaborator prediction, prior work selection, contribution generation, and impact prediction. PreScience is a carefully curated dataset of 98K recent AI-related research papers, featuring disambiguated author identities, temporally aligned scholarly metadata, and a structured graph of companion author publication histories and citations spanning 502K total papers. We develop baselines and evaluations for each task, including LACERScore, a novel ...
