Large Language Models (LLMs) have been extensively applied in various recommendation scenarios, including bundle generation, thanks to their exceptional reasoning capabilities and comprehensive knowledge. However, exploiting large-scale LLMs for bundle generation introduces significant efficiency challenges—primarily high computational costs during fine-tuning and inference due to their massive parameterization. Knowledge Distillation (KD) offers a promising solution by transferring expertise from large teacher models to more compact student models. This study systematically investigates KD approaches for bundle generation with the goal of minimizing computational demands while preserving performance. Specifically, we explore three critical research questions: (1) how does the format of distilled knowledge impact bundle generation performance? (2) to what extent does the quantity of distilled knowledge influence the performance? and (3) how do different ways of utilizing the distilled knowledge affect the performance? To support this investigation, we propose a comprehensive KD framework that (i) progressively extracts knowledge from raw data in increasingly complex forms, i.e., frequent patterns \(\rightarrow\) formalized rules \(\rightarrow\) deep thoughts; (ii) captures varying quantities of distilled knowledge through different sampling strategies, multi-domain accumulation, and multi-format aggregation; and (iii) exploits complementary LLM adaptation techniques—in-context learning, supervised fine-tuning, and their combination—to leverage the distilled knowledge for domain-specific adaptation and enhanced efficiency in small student models. Through extensive experiments on multiple real-world datasets, we provide valuable insights into how knowledge format, quantity, and utilization methods collectively shape the performance of LLM-based bundle generation, which exhibits the significant potential of KD for more efficient yet effective LLM-based bundle generation.

Contemporary Vision-Language Models (VLMs) such as CLIP offer an attractive zero-shot classification functionality facilitated by large-scale vision-language pre-training. However, they remain vulnerable to adversarial attacks, a critical security threat in realistic deployment. Adversarially robust fine-tuning provides generalizable robustness on new datasets while preserving natural performance by fine-tuning the pre-trained models. Fine-tuning robust CLIP typically relies on adversaries generated solely from the vision branch. However, this singular focus on the vision modality, coupled with static text prompts used as fixed category prototypes, limits the robustness achieved through dual-modality fine-tuning. We observe for CLIP fine-tuning that zero-shot adversarial robustness improves when we (i) stabilize the modality gap (a phenomenon where image and text features occupy different feature space regions) and (ii) lower/stabilize gradient norms. Both these steps enjoy further improvement of robustness if one fine-tunes with both visual and text adversaries. For both modalities, we leverage (i) the maximization of an effective rank of features and (ii) noise modulation of features. We show that maximizing the effective rank helps lower and stabilize the modality gap over adversaries with varying perturbation radii. The noise modulation of features, achieved by the so-called count sketching, lowers/stabilizes gradient norms. We outperform the state of the art on 15 datasets. We provide the first insights into the effects of modality gap & gradient norms in VLM fine-tuning.

Most existing bundle generation approaches fall short in generating fixed-size bundles. Furthermore, they often neglect the underlying user intents reflected by the bundles in the generation process, resulting in less intelligible bundles. This paper addresses these limitations through the exploration of two interrelated tasks, i.e., personalized bundle generation and the underlying intent inference, based on different user sessions. Inspired by the reasoning capabilities of large language models (LLMs), we propose an adaptive in-context learning paradigm, which allows LLMs to draw tailored lessons from related sessions as demonstrations, enhancing the performance on target sessions. Specifically, we first employ retrieval augmented generation to identify nearest neighbor sessions, and then carefully design prompts to guide LLMs in executing both tasks on these neighbor sessions. To tackle reliability and hallucination challenges, we further introduce (1) a self-correction strategy promoting mutual improvements of the two tasks without supervision signals and (2) an auto-feedback mechanism for adaptive supervision based on the distinct mistakes made by LLMs on different neighbor sessions. Thereby, the target session can gain customized lessons for improved performance by observing the demonstrations of its neighbor sessions. Experiments on three real-world datasets demonstrate the effectiveness of our proposed method.

The goal of intent-aware session recommendation (ISR) approaches is to capture user intents within a session for accurate next-item prediction. However, the capability of these approaches is limited by assuming all sessions have a uniform and fixed number of intents. In reality, user sessions can vary, where the number of intentions may differ from one to another. Moreover, they can only learn user intents in the latent space, which further restricts the model's transparency. To ease these issues, we propose a simple yet effective paradigm for ISR motivated by the advanced reasoning capability of large language models (LLMs). Specifically, we first create an initial prompt to instruct LLMs to predict the next item by inferring varying user intents reflected in a session. Then, we propose an effective optimization mechanism to automatically optimize prompts with an iterative self-reflection. Finally, we leverage the robust generalizability of LLMs across diverse domains to efficiently select the optimal prompt for ISR. As such, the proposed paradigm effectively guides LLMs to identify varying user intents at a semantic level, thus delivering more accurate and comprehensible recommendations. Extensive experiments on three real-world datasets verify the superiority of our proposed method.

Product bundling represents a prevalent marketing strategy in both offline stores and e-commerce systems. Despite its widespread use, previous studies on bundle recommendation face two significant limitations. Firstly, they rely on noisy datasets, where bundles are defined by heuristics, e.g., products co-purchased in the same session. Secondly, they target specific tasks by holding unrealistic assumptions, e.g., the availability of bundles for recommendation directly. This paper proposes to take a step back and considers the process of bundle recommendation from a holistic user experience perspective. We first construct high-quality bundle datasets with rich metadata, particularly bundle intents, through a carefully designed crowd-sourcing task. We then define a series of tasks that together, support all key steps in a typical bundle recommendation process, from bundle detection, completion and ranking, to explanation and auto-naming, whereby 19 research questions are raised correspondingly to guide the analysis. Finally, we conduct extensive experiments and analyses with representative recommendation models and large language models (LLMs), demonstrating the challenges and opportunities, especially with the emergence of LLMs. To summarize, our study contributes by introducing novel data sources, paving the way for new research avenues, and offering insights to guide product bundling in real e-commerce platforms.

Conversational recommender systems (CRSs) have become crucial emerging research topics in the field of RSs, thanks to their natural advantages of explicitly acquiring user preferences via interactive conversations and revealing the reasons behind recommendations. However, the majority of current CRSs are text-based, which is less user-friendly and may pose challenges for certain users, such as those with visual impairments or limited writing and reading abilities. Therefore,for the first time, this paper investigates the potential of voice-based CRS (VCRSs) to revolutionize the way users interact with RSs in a natural, intuitive, convenient, and accessible fashion. To support such studies, we create two VCRSs benchmark datasets in the e-commerce and movie domains, after realizing the lack of such datasets through an exhaustive literature review. Specifically, we first empirically verify the benefits and necessity of creating such datasets. Thereafter, we convert the user-item interactions to text-based conversations through the ChatGPT-driven prompts for generating diverse and natural templates, and then synthesize the corresponding audios via the text-to-speech model. Meanwhile, a number of strategies are delicately designed to ensure the naturalness and high quality of voice conversations. On this basis, we further explore the potential solutions and point out possible directions to build end-to-end VCRSs by seamlessly extracting and integrating voice-based inputs, thus delivering performance-enhanced, self-explainable, and user-friendly VCRSs. Our study aims to establish the foundation and motivate further pioneering research in the emerging field of VCRSs. This aligns with the principles of explainable AI and AI for social good, viz., utilizing technology's potential to create a fair, sustainable, and just world. Our codes and datasets are available on GitHub (https://github.com/hyllll/VCRS ).

In this paper, we work on a challenging dynamic transfer regression problem where domains come in a streaming manner. At each time stage, a new domain emerges and is taken as the target domain while all the domains in previous time stages are taken as source domains. We propose a transfer Gaussian process model GPdk with a novel dynamic transfer kernel DyTK to handle the dynamic transfer regression problem. Specifically, DyTK is with a sequential form to fit the domain stream. To adaptively control the knowledge transfer strength, DyTK is designed to be capable of modeling the inter-domain relatedness of every inter-domain pair. A theorem that ensures DyTK to be positive semi-definite is then proposed. We also theoretically analyze the transfer performance of GPdk by deriving its generalization error bounds. The error bounds further motivate us to propose a parameter reuse strategy to alleviate the scalability issue of GPdk along time. Extensive experiments on both synthetic and real-world datasets show the effectiveness of GPdk in handling dynamic transfer regression problems.

Policy distillation (PD) has been widely studied in deep reinforcement learning (RL), while existing PD approaches assume that the demonstration data (i.e., state-action pairs in frames) in a decision making sequence is uniformly distributed. This may bring in unwanted bias since RL is a reward maximizing process instead of simple label matching. Given such an issue, we denote the frame importance as its contribution to the expected reward on a particular frame, and hypothesize that adapting such frame importance could benefit the performance of the distilled student policy. To verify our hypothesis, we analyze why and how frame importance matters in RL settings. Based on the analysis, we propose an importance prioritized PD framework that highlights the training on important frames, so as to learn efficiently. Particularly, the frame importance is measured by the reciprocal of weighted Shannon entropy from a teacher policy's action prescriptions. Experiments on Atari games and policy compression tasks show that capturing the frame importance significantly boosts the performance of the distilled policies.

Adversarial examples in automatic speech recognition (ASR) are naturally sounded by humans yet capable of fooling well trained ASR models to transcribe incorrectly. Existing audio adversarial examples are typically constructed by adding constrained perturbations on benign audio inputs. Such attacks are therefore generated with an audio dependent assumption. For the first time, we propose the Speech Synthesising based Attack (SSA), a novel threat model that constructs audio adversarial examples entirely from scratch, i.e., without depending on any existing audio to fool cutting-edge ASR models. To this end, we introduce a conditional variational auto-encoder (CVAE) as the speech synthesiser. Meanwhile, an adaptive sign gradient descent algorithm is proposed to solve the adversarial audio synthesis task. Experiments on three datasets (i.e., Audio Mnist, Common Voice, and Librispeech) show that our method could synthesise naturally sounded audio adversarial examples to mislead the start-of-the-art ASR models. Our web-page containing generated audio demos is at https://sites.google.com/view/ssa-asr/home.

Product bundling is a commonly-used marketing strategy in both offline retailers and online e-commerce systems. Current research on bundle recommendation is limited by: (1) noisy datasets, where bundles are defined by heuristics, e.g., products co-purchased in the same session; and (2) specific tasks, holding unrealistic assumptions, e.g., the availability of bundles for recommendation directly. In this paper, we propose to take a step back and consider the process of bundle recommendation from a holistic user experience perspective. We first construct high-quality bundle datasets with rich meta information, particularly bundle intents, through a carefully designed crowd-sourcing task. We then define a series of tasks that together, support all key steps in a typical bundle recommendation process, from bundle detection, completion, ranking, to explanation and auto-naming. Finally, we conduct extensive experiments and in-depth analysis that demonstrate the challenges of bundle recommendation, arising from the need for capturing complex relations among users, products, and bundles, as well as the research opportunities, especially in graph-based neural methods. To sum up, our study delivers new data sources, opens up new research directions, and provides useful guidance for product bundling in real e-commerce platforms. Our datasets are available at GitHub (\urlhttps://github.com/BundleRec/bundle_recommendation ).

With tremendous amount of recommendation algorithms proposed every year, one critical issue has attracted a considerable amount of attention: there are no effective benchmarks for evaluation, which leads to two major concerns, i.e., unreproducible evaluation and unfair comparison. This paper aims to conduct rigorous (i.e., reproducible and fair) evaluation for implicit-feedback based top-N recommendation algorithms. We first systematically review 85 recommendation papers published at eight top-tier conferences (e.g., RecSys, SIGIR) to summarize important evaluation factors, e.g., data splitting and parameter tuning strategies, etc. Through a holistic empirical study, the impacts of different factors on recommendation performance are then analyzed in-depth. Following that, we create benchmarks with standardized procedures and provide the performance of seven well-tuned state-of-the-arts across six metrics on six widely-used datasets as a reference for later study. Additionally, we release a user-friendly Python toolkit, which differs from existing ones in addressing the broad scope of rigorous evaluation for recommendation. Overall, our work sheds light on the issues in recommendation evaluation and lays the foundation for further investigation. Our code and datasets are available at GitHub (https://github.com/AmazingDD/daisyRec).