LATIDIA · Investigación

OmicsBench: un punto de referencia de múltiples giros a largo plazo de agentes de IA en bioinformática a escala de estudio de investigación

arXiv: 2610.21074v1Tipo de anuncio: nuevo Resumen: Los biólogos computacionales delegan cada vez más análisis sustanciales a los agentes de IA en sesiones de largo horizonte y múltiples turnos, pero la mayoría de los puntos de referencia de la capacidad bioinformática de los agentes

La noticia

arXiv:2610.21074v1 Announce Type: new Abstract: Computational biologists increasingly delegate substantial analyses to AI agents in long-horizon, multi-turn sessions, yet most benchmarks of agents' bioinformatics capabilities pose a single short task with a single-shot prompt. We present OmicsBench, a long-horizon, multi-turn bioinformatics benchmark for AI agents. Each of its 97 tasks links dependent subtasks into a directed acyclic graph, for 662 underlying subtasks in all. Each subtask is posed as one user turn and graded deterministically against an answer key, with optional in-context feedback on incorrect answers. Across 12 models on three harnesses, the slowest agent averages 3.6 hours per task, and no combination comes near saturating the benchmark, with mean