When AI Does Science: Evaluating the Autonomous AI Scientist KOSMOS in Radiation Biology
What this work does
Reconstructs three Kosmos radiobiology analyses and compares proposed gene signatures with random-gene or permutation null models.
Connection to discovery
A concrete example of claim-level discovery validation and of retaining negative findings rather than rewarding persuasive hypotheses.
Evidence
Inspected the statistical audit protocol and the very small-sample DDR example where a seemingly meaningful relationship does not beat the null.
Scope and limitations
Three retrospective cases are not a representative benchmark of the whole system. Small samples and analytic choices limit both positive and negative conclusions; independent reproduction remains necessary.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.