Tool learning, which allows Large Language Models (LLMs) to leverage external tools for solving complex user tasks, has emerged as a promising avenue for extending model capabilities. However, existing approaches primarily focus on data synthesis for fine-tuning LLMs to invoke tools effectively, largely ignoring how to fully stimulate the potential of the model. In this paper, we propose ToolACE-R, a novel framework that includes both model-aware iterative training and adaptive refinement for tool learning. ToolACE-R features a model-aware iterative training procedure that progressively adjust training samples based on the model’s evolving capabilities to maximize its potential. Additionally, it incorporates self-refinement training corpus which emphasizes LLM's ability to iteratively refine their tool calls, optimizing performance without requiring external feedback. Furthermore, we introduce adaptive self-refinement for efficient test-time scaling, where the trained model can autonomously determine when to stop the process based on iterative self-refinement. We conduct extensive experiments across several benchmark datasets, showing that ToolACE-R achieves competitive performance compared to advanced LLMs. The performance can be further improved efficiently through adaptive self-refinement. These results highlight the effectiveness and generalizability of ToolACE-R, offering a promising direction for more efficient and scalable tool learning.

Yirong Zeng, Xiao Ding, Yuxian Wang, Weiwen Liu, Yutai Hou, Wu Ning, Xu Huang, Duyu Tang, Dandan Tu, Bing Qin, Ting Liu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Function calling significantly extends the application boundary of large language models (LLMs), where high-quality and diverse training data is critical for unlocking this capability. However, collecting and annotating real function-calling data is challenging, while synthetic data from existing pipelines often lack coverage and accuracy. In this paper, we present ToolACE, an automatic agentic pipeline designed to generate accurate, complex, and diverse tool-learning data, specifically tailored to the capabilities of LLMs. ToolACE leverages a novel self-evolution synthesis process to curate a comprehensive API pool of 26,507 diverse APIs. Dialogs are further generated through the interplay among multiple agents, under the guidance of a complexity evaluator. To ensure data accuracy, we implement a dual-layer verification system combining rule-based and model-based checks. We demonstrate that models trained on our synthesized data---even with only 8B parameters---achieve state-of-the-art performance, comparable to the latest GPT-4 models. Our model and a subset of the data are publicly available at https://huggingface.co/Team-ACE.

Recommender models capture ever-changing user preferences by training with in-domain user behavior data. These models are typically lightweight, facilitating real-time and large-scale online services. However, these models often falter when tasked with providing more sophisticated functionalities, such as offering explanations or engaging in conversations. Recently, large language models (LLMs) have emerged as a significant advancement towards artificial general intelligence, demonstrating impressive capabilities in instruction comprehension, reasoning, and human interaction. Unfortunately, LLMs lack the understanding of domain-specific item catalogs and behavioral patterns, especially in areas that deviate from general world knowledge, such as online e-commerce. This limitation makes them unsuitable to function as recommender models directly. In this article, we bridge the gap between recommender models and LLMs, combining their respective strengths to create an interactive recommender system. We present an efficient framework, termed as InteRecAgent , which utilizes LLMs as the brain and recommender models as instrumental tools. We first outline a minimal set of essential tools required to transform LLMs into InteRecAgent. To overcome specific challenges associated with LLM-based agents for recommender systems, we enhance three core components, covering memory mechanism, task planning, and tool learning abilities. The InteRecAgent empowers traditional recommender systems, like ID-based matrix factorization models, to evolve into versatile and interactive systems with a natural language interface through the integration of LLMs. Experimental results derived from three public datasets demonstrate that the InteRecAgent delivers strong performance as a conversational recommender system, surpassing general LLMs such as GPT-4.

To protect privacy and meet legal regulations, federated learning (FL) has gained significant attention for training speech-to-text (S2T) systems, including automatic speech recognition (ASR) and speech translation (ST). However, the commonly used FL approach (i.e., FEDAVG) in S2T tasks typically suffers from extensive communication overhead due to multi-round interactions based on the whole model and performance degradation caused by data heterogeneity among clients. To address these issues, we propose a personalized federated S2T framework that introduces FEDLORA, a lightweight LoRA module for client-side tuning and interaction with the server to minimize communication overhead, and FEDMEM, a global model equipped with a k-nearest-neighbor (kNN) classifier that captures client-specific distributional shifts to achieve personalization and overcome data heterogeneity. Extensive experiments based on Conformer and Whisper backbone models on CoVoST and GigaSpeech benchmarks show that our approach significantly reduces the communication overhead on all S2T tasks and effectively personalizes the global model to overcome data heterogeneity.

The weighted squared loss is a common component in several Collaborative Filtering (CF) algorithms for item recommendation, including the representative implicit Alternating Least Squares (iALS). Despite its widespread use, this loss function lacks a clear connection to ranking objectives such as Discounted Cumulative Gain (DCG), posing a fundamental challenge in explaining the exceptional ranking performance observed in these algorithms. In this work, we make a breakthrough by establishing a connection between squared loss and ranking metrics through a Taylor expansion of the DCG-consistent surrogate loss—softmax loss. We also discover a new surrogate squared loss function, namely Ranking-Generalizable Squared (RG$^2$) loss, and conduct thorough theoretical analyses on the DCG-consistency of the proposed loss function. Later, we present an example of utilizing the RG$^2$ loss with Matrix Factorization (MF), coupled with a generalization upper bound and an ALS optimization algorithm that leverages closed-form solutions over all items. Experimental results over three public datasets demonstrate the effectiveness of the RG$^2$ loss, exhibiting ranking performance on par with, or even surpassing, the softmax loss while achieving faster convergence.

Recommender systems are widely used in online services, with embedding-based models being particularly popular due to their expressiveness in representing complex signals. However, these models often function as a black box, making them less transparent and reliable for both users and developers. Recently, large language models (LLMs) have demonstrated remarkable intelligence in understanding, reasoning, and instruction following. This paper presents the initial exploration of using LLMs as surrogate models to explaining black-box recommender models. The primary concept involves training LLMs to comprehend and emulate the behavior of target recommender models. By leveraging LLMs' own extensive world knowledge and multi-step reasoning abilities, these aligned LLMs can serve as advanced surrogates, capable of reasoning about observations. Moreover, employing natural language as an interface allows for the creation of customizable explanations that can be adapted to individual user preferences. To facilitate an effective alignment, we introduce three methods: behavior alignment, intention alignment, and hybrid alignment. Behavior alignment operates in the language space, representing user preferences and item information as text to mimic the target model's behavior; intention alignment works in the latent space of the recommendation model, using user and item representations to understand the model's behavior; hybrid alignment combines both language and latent spaces. Comprehensive experiments conducted on three public datasets show that our approach yields promising results in understanding and mimicking target models, producing high-quality, high-fidelity, and distinct explanations. Our code is available at https://github.com/microsoft/RecAI.

Recommendation systems guide users in locating their desired information within extensive content repositories. Usually, a recommendation model is optimized to enhance accuracy metrics from a user utility standpoint, such as click-through rate or matching relevance. However, a responsible industrial recommendation model must address not only user utility (responsibility to users) but also other objectives, including increasing platform revenue (responsibility to platforms), ensuring fairness (responsibility to content creators), and maintaining unbiasedness (responsibility to long-term healthy development). Multi-objective learning is a promising approach for achieving responsible recommendation models. Nevertheless, current methods encounter two challenges: difficulty in scaling to heterogeneous objectives within a unified framework, and inadequate controllability over objective priority during optimization, leading to uncontrollable solutions.

Abstract The advent of large language models marks a revolutionary breakthrough in artificial intelligence. With the unprecedented scale of training and model parameters, the capability of large language models has been dramatically improved, leading to human-like performances in understanding, language synthesizing, common-sense reasoning, etc. Such a major leap forward in general AI capacity will fundamentally change the pattern of how personalization is conducted. For one thing, it will reform the way of interaction between humans and personalization systems. Instead of being a passive medium of information filtering, like conventional recommender systems and search engines, large language models present the foundation for active user engagement. On top of such a new foundation, users’ requests can be proactively explored, and users’ required information can be delivered in a natural, interactable, and explainable way. For another thing, it will also considerably expand the scope of personalization, making it grow from the sole function of collecting personalized information to the compound function of providing personalized services. By leveraging large language models as a general-purpose interface, the personalization systems may compile user’s requests into plans, calls the functions of external tools (e.g., search engines, calculators, service APIs, etc.) to execute the plans, and integrate the tools’ outputs to complete the end-to-end personalization tasks. Today, large language models are still being rapidly developed, whereas the application in personalization is largely unexplored. Therefore, we consider it to be right the time to review the challenges in personalization and the opportunities to address them with large language models. In particular, we dedicate this perspective paper to the discussion of the following aspects: the development and challenges for the existing personalization system, the newly emerged capabilities of large language models, and the potential ways of making use of large language models for personalization.

A dozen recommendation libraries have recently been developed to accommodate popular recommendation algorithms for reproducibility. However, they are almost simply a collection of algorithms, overlooking the modularization of recommendation algorithms and their usage in practical scenarios. Algorithmic modularization has the following advantages: 1) helps to understand the effectiveness of each algorithm; 2) easily assembles new algorithms with well-performed modules by either drag-and-drop programming or automatic machine learning; 3) enables reinforcement between algorithms since one algorithm may act as a module of another algorithm. To this end, we develop a highly-modularized recommender system -- RecStudio, in which any recommendation algorithm is categorized into either a ranker or a retriever. In the RecStudio library, we implement 90 recommendation algorithms with the pure Pytorch, covering both common algorithms in other libraries and complex algorithms involving multiple recommendation models. RecStudio is featured from several perspectives, such as index-supported efficient recommendation and evaluation, GPU-accelerated negative sampling, hyperparameter learning on the validation, and cooperation between the retriever and ranker. RecStudio is also equipped with a web service, where the recommendation pipeline can be quickly established and visually evaluated on selected datasets, and the evaluation results are automatically archived and visualized in a leaderboard. The project and documents are released at http://recstudio.org.cn.

Deep recommender systems (DRS) are intensively applied in modern web services. To deal with the massive web contents, DRS employs a two-stage workflow: retrieval and ranking, to generate its recommendation results. The retriever aims to select a small set of relevant candidates from the entire items with high efficiency; while the ranker, usually more precise but time-consuming, is supposed to further refine the best items from the retrieved candidates. Traditionally, the two components are trained either independently or within a simple cascading pipeline, which is prone to poor collaboration effect. Though some latest works suggested to train retriever and ranker jointly, there still exist many severe limitations: item distribution shift between training and inference, false negative, and misalignment of ranking order. As such, it remains to explore effective collaborations between retriever and ranker.

Variational AutoEncoder (VAE) has been extended as a representative nonlinear method for collaborative filtering. However, the bottleneck of VAE lies in the softmax computation over all items, such that it takes linear costs in the number of items to compute the loss and gradient for optimization. This hinders the practical use due to millions of items in real-world scenarios. Importance sampling is an effective approximation method, based on which the sampled softmax has been derived. However, existing methods usually exploit the uniform or popularity sampler as proposal distributions, leading to a large bias of gradient estimation. To this end, we propose to decompose the inner-product-based softmax probability based on the inverted multi-index, leading to sublinear-time and highly accurate sampling. Based on the proposed proposals, we develop a fast Variational AutoEncoder (FastVAE) for collaborative filtering. FastVAE can outperform the state-of-the-art baselines in terms of both sampling quality and efficiency according to the experiments on three real-world datasets.