Newest papers

CADS: Measuring the Incremental Contribution of LLM Proposers in Sequential Biological Screens

Guanxing Chen, Nihui Shao, Haohuai He, Zhenchao Tang, Pengwei Hu, Zhi-An Huang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.24597 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Los modelos de intervalos de lenguaje adecuados se utilizan cada vez más para guiar las pantallas biológicas secuenciales, sin embargo, las evaluaciones estándar se basan en los contadores crudos que confunden el valor de un proponente con la capacidad de descubrimiento del bucle algorítmico subyacente. El muestreo adaptativo y las lecturas experimentales corrompidas también dejan comparaciones de retroalimentación estándar sin una garantía de falso descubrimiento. Presentamos las pantallas de descubrimiento de agentes certificados (CADS), un marco de medición controlado por errores que cuantifica el valor incremental de un proponente contra un bucle libre de LLM idéntico. Al integrar los límites del ratio de probabilidad, los procesos electrónicos y la familia completa de e-BH, CADS controla la tasa

Machine translation (Meta M2M100); the English original follows.

Abstract

Large language models are increasingly used to guide sequential biological screens, yet standard evaluations rely on raw hit counts that conflate a proposer's value with the discovery capacity of the underlying algorithmic harness. Adaptive sampling and corrupted experimental readouts also leave standard feedback comparisons without a false-discovery guarantee. We present Certified Agentic Discovery Screens (CADS), an error-controlled measurement framework that quantifies a proposer's incremental value against an identical LLM-free loop. By integrating capped likelihood-ratio bounds, e-processes, and complete-family e-BH, CADS controls the terminal false discovery rate under the stated null and corruption conditions despite arbitrary dependence, adaptive sampling, and optional stopping. In DepMap-backed simulations, CADS exposes two evaluation artifacts: the LLM-free harness alone clears the feedback gate, and raw origin counts overstate LLM contributions because LLM picks displace fallback discoveries. On the initial 20 seeds, supplying tested-candidate history to Qwen2.5-7B shifts its net contribution from +0.35, with a 95% interval covering zero, to +2.90, with the interval above zero. This initial net effect did not persist in a 50-seed matched-quota replication. These results support net gain over an LLM-free loop, measured between separately certified sets, as the appropriate measure of proposer contribution.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Measure (data warehouse), Interval (graph theory), Value (mathematics), Conflation, Certification

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange