Examples
Finished analyses that anyone can open. Each map shows a corpus’s contribution claims as dots, coloured by cluster; open one to see how its clusters grow and shrink year by year.
-
ACL venues 2018–2026 (Drift Inspector original)
15,457 papers 2018–2026
Across six ACL venues, 2018–2026: efficient reasoning in LLMs rises 16 points of paper share, while word and sentence embeddings fall 13.
-
AI for Science workshops 2021–2025
680 papers 2021–2025
Scientific reasoning with LLMs and physics-informed neural operators grow fastest across 2021–2025, while data-efficient pretraining fades.
-
NeurIPS 2024–2025 main track, ML-for-science areas
556 papers 2024–2025
Medical imaging foundation models gain the most share from 2024 to 2025; tabular deep learning and neural weather forecasting shrink.
-
AI for science: KDD 2026 vs NeurIPS 2024–2025 vs AI4Science workshops
Comparison of 3 analyses 1,442 papers 2021–2026
KDD 2026 over-indexes on scientific reasoning agents and RL for discovery, NeurIPS on LLM theorem proving, the workshops on cosmological simulation-based inference.