PulseAugur
EN
LIVE 07:30:16

New PORTS method optimizes LLM tool selection with preference learning

By PulseAugur Editorial · [2 sources] ·

Researchers have developed PORTS, a new method for training retrievers to better select external tools for large language models (LLMs). This approach uses a preference optimization technique that aligns the retriever with the LLM's needs by optimizing the correlation between tool selection probabilities and downstream performance. PORTS also enforces a contrastive semantic loss between documentation strings, improving the retriever's ability to find helpful tools. Experiments across multiple datasets, encoder models, and LLMs demonstrate PORTS's versatility and significant improvement in tool selection accuracy, with low computational demands allowing for generalization to new queries and tools. AI

IMPACT Enhances LLM capabilities by improving the selection of external tools, potentially leading to more efficient and accurate task completion.

RANK_REASON The cluster contains an academic paper detailing a new method for improving LLM tool selection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New PORTS method optimizes LLM tool selection with preference learning

COVERAGE [2]

  1. arXiv cs.AI TIER_1 English(EN) · Lorenzo Molfetta, Giacomo Frisoni, Nicol\`o Monaldini, Gianluca Moro ·

    PORTS: Preference-Optimized Retrievers for Tool Selection with Large Language Models

    arXiv:2607.05441v1 Announce Type: cross Abstract: Integrating external tools with Large Language Models (LLMs) has emerged as a promising paradigm for accomplishing complex tasks. Since LLMs still struggle to effectively manage large tool collections, researchers have begun explo…

  2. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Gianluca Moro ·

    PORTS: Preference-Optimized Retrievers for Tool Selection with Large Language Models

    Integrating external tools with Large Language Models (LLMs) has emerged as a promising paradigm for accomplishing complex tasks. Since LLMs still struggle to effectively manage large tool collections, researchers have begun exploring retrieval-based methods to pre-select the mos…

RELATED ENTITIES

RELATED TOPICS