Latent Preference Modeling for Cross-Session Personalized Tool Calling

Hugging Face Daily Papers Papers
personalized-tool-calling llm-agents memory-augmentation user-preferences cross-session

Summary

Introduces MPT benchmark and PRefine method for cross-session personalized tool calling that captures user choice reasoning with minimal token overhead.

Users often omit essential details in their requests to LLM-based agents, resulting in under-specified inputs for tool use. This poses a fundamental challenge for tool-augmented agents, as API execution typically requires complete arguments, highlighting the need for personalized tool calling. To study this problem, we introduce MPT, a benchmark comprising 265 multi-session dialogues that cover three challenges: Preference Recall, Preference Induction, and Preference Transfer. We also propose PRefine, a test-time memory-augmented method that represents user preferences as evolving hypotheses. Through a generate--verify--refine loop, it extracts reusable constraints from history and improves tool-calling accuracy while using only 1.24% of the tokens required by full-history prompting. These results indicate that robust personalization in agentic systems depends on memory that captures the reasons behind user choices, not just the choices themselves.
Original Article Export to Word Export to PDF
View Cached Full Text

Cached at: 04/21/26, 11:27 AM

Paper page - Latent Preference Modeling for Cross-Session Personalized Tool Calling

Source: https://huggingface.co/papers/2604.17886

Abstract

Personalized tool calling in LLM-based agents is improved through memory-augmented methods that capture user choice reasoning rather than just choices, using minimal token overhead.

Users often omit essential details in their requests to LLM-based agents, resulting in under-specified inputs for tool use. This poses a fundamental challenge fortool-augmented agents, asAPI executiontypically requires complete arguments, highlighting the need forpersonalized tool calling. To study this problem, we introduce MPT, a benchmark comprising 265multi-session dialoguesthat cover three challenges:Preference Recall,Preference Induction, andPreference Transfer. We also proposePRefine, a test-time memory-augmented method that representsuser preferencesas evolving hypotheses. Through agenerate--verify--refine loop, it extracts reusable constraints from history and improves tool-calling accuracy while using only 1.24% of the tokens required by full-history prompting. These results indicate that robust personalization in agentic systems depends on memory that captures the reasons behind user choices, not just the choices themselves.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2604\.17886

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2604.17886 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2604.17886 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2604.17886 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

FSPO: Few-Shot Optimization of Synthetic Preferences Personalizes to Real Users

arXiv cs.CL

FSPO proposes a few-shot preference optimization algorithm for LLM personalization that reframes reward modeling as meta-learning, enabling models to quickly infer personalized reward functions from limited user preferences. The method achieves 87% personalization performance on synthetic users and 70% on real users through careful synthetic preference dataset construction.

PersonaVLM: Long-Term Personalized Multimodal LLMs

Hugging Face Daily Papers

PersonaVLM introduces a personalized multimodal LLM framework that enables long-term user adaptation through memory retention, multi-turn reasoning, and response alignment, outperforming GPT-4o by 5.2% on the new Persona-MME benchmark.

Preference Estimation via Opponent Modeling in Multi-Agent Negotiation

arXiv cs.CL

This paper proposes a novel preference estimation method that integrates natural language information from LLMs into a structured Bayesian opponent modeling framework for multi-agent negotiation. The approach leverages LLMs to extract qualitative cues from utterances and convert them into probabilistic formats, demonstrating improved agreement rates and preference estimation accuracy on multi-party negotiation benchmarks.

IPQA: A Benchmark for Core Intent Identification in Personalized Question Answering

arXiv cs.CL

IPQA introduces a benchmark for evaluating core intent identification in personalized question answering, addressing a gap in existing metrics that focus on response quality rather than intent understanding. The paper presents a dataset construction methodology grounded in bounded rationality and demonstrates that state-of-the-art language models struggle with identifying user-prioritized intents from answer selection patterns.

Inference-Time Budget Control for LLM Search Agents

arXiv cs.AI

This paper introduces a two-stage inference-time budget control method for LLM search agents, using Value-of-Information scores to optimize tool-call and token allocation during multi-hop question answering.