Experimental Attempts in Electronic Lab Notebooks: A Dataset Proposal for Scientific Debugging

Modern science has a publication bias: successful protocols, curated papers, and cleaned-up scientific narratives are often overrepresented. However, scientific AI agents must also reason about failed runs, deviations, and troubleshooting under incomplete and noisy evidence. Such events are often recorded in laboratory notebooks, yet public datasets of notebook-derived experimental attempts remain scarce despite the growth of electronic laboratory notebooks (ELNs) and open-notebook practices. We propose a dataset of run-level experimental attempt records with fields such as goal, hypothesis when available, run text, observations, status, deviation, response, and evidence spans for inferred labels.