Mapping the lifecycle of scientific contributions: From generation and evaluation to impact diffusion

Abstract Understanding how scientific contributions evolve from their inception to long‐term influence is central to the study of knowledge dynamics and research evaluation. This study investigates the evolution of actors' cognitive perceptions of scientific contributions in academic papers by harnessing large language models (LLMs) to conduct a combined analysis of full texts, peer review comments, and citation contexts of over 29,000 articles from PLOS ONE . Drawing on the social construction theory of scientific knowledge, we propose a dynamic construction lifecycle comprising three stages—generation, external evaluation, and impact diffusion—shaped respectively by authors, reviewers, and citers. By classifying contribution types from several textual sources, we examined their consistency and evolution in terms of both type and strength across lifecycle stages. The results reveal that authors highlight experimental and application‐oriented contributions; reviewers focus on conceptual dimensions and apply stricter criteria; citers emphasize experimental and methodological aspects. Furthermore, a cognitive gap emerges between reviewers and other actors, indicating that reviewers often rely on perceived significance rather than contribution typologies. In contrast, alignment between authors and citers suggests that citation‐based indicators, despite limitations, capture meaningful dimensions of scholarly recognition. Analysis of contribution evolution patterns shows both divergent and stable trajectories with non‐linear fluctuations in perceived strength.