Wei Guo, Yuanjian Xu, Jianing Hao, Yongyao Duan, Ran Chen, Zhong Li, Guang Zhang · 2026 · arXiv
DOI: 10.48550/arXiv.2610.29751 · open-access PDF
Model bahasa besar (LLM) semakin diimplementasikan sebagai agen, namun tidak jelas bagaimana kemampuan dasar mereka diterjemahkan menjadi kemampuan agen. skor end-to-end meninggalkan tidak terpecahkan apakah kemampuan benchmark rendah atau pengimplementasian yang lemah mendorong kegagalan, membatasi diagnosis yang dapat dilakukan. Kami menghubungkan kemampuan dasar dan agen melalui kerangka pengukuran tingkat operasi. Kami mewakili setiap model dengan profil dasar yang berasal dari referensi kemampuan dan profil operasi yang berasal dari trajektori interaksi, dan mendefinisikan kesenjangan realisasi kemampuan sebagai penyimpangan
Machine translation (Meta M2M100); the English original follows.
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
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.29751, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange