Seungone Kim

@seungonekim

Ph.D. student @LTIatCMU working on AI for science & LLM evals & nuclear fusion | Prev: @AIatMeta (FAIR) @kaist_ai @yonsei_u

following
followers

Most of 135 agent-produced papers reportedly fail integrity checks in a new benchmark

A post describes a benchmark testing nine frontier agents on 15 open scientific follow-up problems developed with domain scientists.

10 Sources, 7d ago, updated 7d ago

AI agents reportedly showed different ‘research taste’ in an ICML 2026 pilot

A researcher says the agents acted as area chairs and ranked all ICML 2026 papers in the pilot.

2 Sources, 16d ago, first seen 16d ago

EurekaBench introduced to assess AI agents’ scientific insight across six domains

A project contributor says the benchmark examines whether agents can discover new insights through experimentation, beyond trial-and-error optimization.

8 Sources, 19d ago, updated 19d ago

ScholarCatalyst focuses on finding papers that inspire new research

Akari Asai describes a benchmark involving hundreds of scientists and room to improve paper discovery.

11 Sources, 18d ago, updated 18d ago

AI Reviewers Work Featured in Science Magazine

Seungone Kim shares feature of AI reviewers work in Science article on peer review.

4 Sources, 49d ago, first seen 49d ago