Newest papers

Bridging Foundational and Agentic Capabilities of LLMs

Wei Guo, Yuanjian Xu, Jianing Hao, Yongyao Duan, Ran Chen, Zhong Li, Guang Zhang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29751 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Os grandes modelos de linguagem (LLMs) são cada vez mais implantados como agentes, mas permanece incerto como suas capacidades fundamentais se traduzem em capacidades de agente. As pontuações de ponta a ponta deixam não resolvidas se a capacidade de referência baixa ou a implantação fraca conduz a um fracasso, limitando o diagnóstico acionável. Nós cruzamos as capacidades fundamentais e de agente através de um quadro de medição de nível de operação. Representamos cada modelo com um perfil de fundação derivado de benchmarks de capacidade e um perfil de operação derivado de trajetórias de interação, e definimos as lacunas de realização de capacidade como desvios do desempenho observado de previsões baseadas em capacidades fundamentais. Experimentos em quinze modelos de serviços em $^2$-

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

Abstract

Large language models (LLMs) are increasingly deployed as agents, yet it remains unclear how their foundational capabilities translate into agentic capabilities. End-to-end scores leave unresolved whether low benchmark capability or weak deployment drives a failure, limiting actionable diagnosis. We bridge foundational and agentic capabilities through an operation-level measurement framework. We represent each model with a foundational profile derived from capability benchmarks and an operation profile derived from interaction trajectories, and define capability realization gaps as deviations of observed operation performance from predictions based on foundational capabilities. Experiments on fifteen model services in $τ^2$-Bench show that foundational profiles only partially predict agentic operations. Models with comparable profiles can differ substantially in execution, and these gaps remain consistent across task subsets. Operation profiles add information and substantially improve task-score prediction. The framework locates execution departures at the operation level to inform model development and agent evaluation. We will release our code and analysis artifacts.

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

Topics: Bridging (networking), Limiting, Computer science, Bridge (graph theory), Software deployment, Task (project management)

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