Newest papers

Training Small Language Models for Transferable Tool Selection

Zhaoyi Zhou, Akshayaa Magesh, Jeongyeol Kwon, Artem Kuriksha, Daniel Jiang, Andrea Zanette, Ayush Jain · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31483 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Como os agentes do LLM operam em uma corpora de mais de dez mil ferramentas, a seleção da ferramenta certa para uma instrução de usuário complexa em vários passos torna-se um ponto central. As descrições de ferramentas não podem todos se encaixar no contexto, e os retrievers baseados em relevância apenas classificam ferramentas sem saber quais são realmente necessárias. O trabalho recente, em vez disso, aproveita os LLMs comerciais proprietários como Claude ou ChatGPT para dirigir um retriever, decompor a instrução e selecionar sobre candidatos recuperados. Mostramos que um pequeno modelo aberto pode ser treinado através de aprendizado de reforço para executar a mesma funcionalidade. No ToolBench, um modelo Qwen3-8B pós-treinado com a nossa receita supera Llama-4-Maverick (402

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

Abstract

As LLM agents operate over corpora of more than ten thousand tools, selecting the right tool for a complex multi-step user instruction becomes a central bottleneck. Tool descriptions cannot all fit in context, and relevance-based retrievers only rank tools without knowing which are actually needed. Recent work instead leverages proprietary commercial LLMs like Claude or ChatGPT to drive a retriever, decomposing the instruction and selecting over retrieved candidates. We show that a small, open model can be trained via reinforcement learning to perform the same functionality. On ToolBench, a Qwen3-8B model post-trained with our recipe outperforms Llama-4-Maverick (402B). In addition, the trained model transfers directly to different retrievers and generalizes to new tool corpora in different domains.

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

Topics: Computer science, Selection (genetic algorithm), Artificial intelligence, Rank (graph theory), Language model, Training (meteorology)

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31483, 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.txt

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