Search papers, labs, and topics across Lattice.
This paper introduces Xcientist, a research harness designed to externalize the synthesis and validation processes in AI-driven scientific workflows, making them more transparent and accountable. By organizing various research artifacts鈥攕uch as literature evidence and implementation plans鈥擷cientist addresses the issue of claim drift, where automated systems produce outputs that deviate from their original claims. The findings demonstrate that maintaining traceable trajectories throughout the research process enhances the reliability and scientific integrity of AI-generated mechanisms across multiple domains, including memory systems and neural networks.
Claim drift in automated research can lead to significant discrepancies, but Xcientist ensures that every generated mechanism remains accountable and traceable back to its evidential roots.
AI systems can increasingly automate scientific workflows, but the reasoning that links prior evidence, generated ideas, experiments and final claims often remains implicit inside model inference. Here we introduce Xcientist, a research harness that externalizes research synthesis and experimental validation into inspectable, contract-governed processes. Xcientist organizes literature evidence, idea states, implementation plans, ablation records and repair traces as persistent research artifacts, so that generated mechanisms can be grounded, executed, tested and revised without losing their evidential basis. We identify claim drift as a failure mode of automated research, where runnable artifacts no longer support the mechanism originally claimed. Across training-free memory systems, graph-structured traffic forecasting and multi-scale physics-informed neural networks, Xcientist preserves traceable trajectories from problem formulation to mechanism design, validation and bounded revision. These results suggest that AI scientists should be evaluated not only by their final artifacts, but by whether their synthesis and validation processes remain attributable, inspectable and scientifically accountable.
