Retrieval-augmented Generation (RAG) integrates Large Language Models (LLMs) with retrievers to access external knowledge, improving the factuality of LLM generation in knowledge-grounded tasks. To optimize the RAG performance, most previous work independently fine-tunes the retriever to adapt to frozen LLMs or trains the LLMs to use documents retrieved by off-the-shelf retrievers, lacking end-to-end training supervision. Recent work addresses this limitation by jointly training these two components but relies on overly simplifying assumptions of document independence, which has been criticized for being far from real-world scenarios. Thus, effectively optimizing the overall RAG performance remains a critical challenge. We propose a Direct Retrieval-augmented Optimization ( DRO ) framework that enables end-to-end training of two key components: (i) a generative knowledge selection model and (ii) an LLM generator. DRO alternates between two phases: (i) document permutation estimation and (ii) re-weighted maximization, progressively improving RAG components through a variational approach. In the estimation step, we treat document permutation as a latent variable and directly estimate its distribution from the selection model by applying an importance sampling strategy. In the maximization step, we calibrate the optimization expectation using importance weights and jointly train the selection model and LLM generator. Our theoretical analysis reveals that DRO is analogous to policy-gradient methods in reinforcement learning. Extensive experiments conducted on five datasets illustrate that DRO outperforms the best baseline with 5–15% improvements in EM and F1. We also qualitatively analyze the stability, convergence, and variance of DRO. (Code is available on DRO GitHub ).

Generative retrieval (GR) has emerged as a promising paradigm in information retrieval (IR). However, most existing GR models are developed and evaluated using a static document collection, and their performance in dynamic corpora where document collections evolve continuously is rarely studied. In this paper, we first reproduce and systematically evaluate various representative GR approaches over dynamic corpora. Through extensive experiments, we reveal that existing GR models with text-based docids show superior generalization to unseen documents. We observe that the more fine-grained the docid design in the GR model, the better its performance over dynamic corpora, surpassing BM25 and even being comparable to dense retrieval methods. While GR models with numeric-based docids show high efficiency, their performance drops significantly over dynamic corpora. Furthermore, our experiments find that the underperformance of numeric-based docids is partly due to their excessive tendency toward the initial document set, which likely results from overfitting on the training set. We then conduct an in-depth analysis of the best-performing GR methods. We identify three critical advantages of text-based docids in dynamic corpora: (i) Semantic alignment with language models' pretrained knowledge (ii) Fine-grained docid design, and (iii) High lexical diversity. Building on these insights, we finally propose a novel multi-docid design that leverages both the efficiency of numeric-based docids and the effectiveness of text-based docids, achieving improved performance in dynamic corpus without requiring additional retraining. Our work offers empirical evidence for advancing GR methods over dynamic corpora and paves the way for developing more generalized yet efficient GR models in real-world search engines.

Large Language Models (LLMs) are increasingly employed in zero-shot documents ranking, yielding commendable results. However, several significant challenges still persist in LLMs for ranking: (1) LLMs are constrained by limited input length, precluding them from processing a large number of documents simultaneously; (2) The output document sequence is influenced by the input order of documents, resulting in inconsistent ranking outcomes; (3) Achieving a balance between cost and ranking performance is challenging. To tackle these issues, we introduce a novel documents ranking method called TourRank1. which is inspired by the sport tournaments, such as FIFA World Cup. Specifically, we 1) overcome the limitation in input length and reduce the ranking latency by incorporating a multi-stage grouping strategy similar to the parallel group stage of sport tournaments; 2) improve the ranking performance and robustness to input orders by using a points system to ensemble multiple ranking results. We test TourRank with different LLMs on the TREC DL datasets and the BEIR benchmark. The experimental results demonstrate that TourRank delivers state-of-the-art performance at a modest cost.

Generative models have emerged as a promising utility to enhance recommender systems. It is essential to model both item content and user-item collaborative interactions in a unified generative framework for better recommendation. Although some existing large language model (LLM)-based methods contribute to fusing content information and collaborative signals, they fundamentally rely on textual language generation, which is not fully aligned with the recommendation task. How to integrate content knowledge and collaborative interaction signals in a generative framework tailored for item recommendation is still an open research challenge.

Task-oriented dialogue systems (TDSs) are assessed mainly in an offline setting or through human evaluation. The evaluation is often limited to single-turn or is very time-intensive. As an alternative, user simulators that mimic user behavior allow us to consider a broad set of user goals to generate human-like conversations for simulated evaluation. Employing existing user simulators to evaluate TDSs is challenging as user simulators are primarily designed to optimize dialogue policies for TDSs and have limited evaluation capabilities. Moreover, the evaluation of user simulators is an open challenge. In this work, we propose a metaphorical user simulator for end-to-end TDS evaluation, where we define a simulator to be metaphorical if it simulates a user’s analogical thinking in interactions with systems. We also propose a tester-based evaluation framework to generate variants, i.e., dialogue systems with different capabilities. Our user simulator constructs a metaphorical user model that assists the simulator in reasoning by referring to prior knowledge when encountering new items. We estimate the quality of simulators by checking the simulated interactions between simulators and variants. Our experiments are conducted using three TDS datasets. The proposed user simulator demonstrates better consistency with manual evaluation than an agenda-based simulator and a seq2seq model on three datasets; our tester framework demonstrates efficiency and has been tested on multiple tasks, such as conversational recommendation and e-commerce dialogues.

Explanations in conventional recommender systems have demonstrated benefits in helping the user understand the rationality of the recommendations and improving the system's efficiency, transparency, and trustworthiness. In the conversational environment, multiple contextualized explanations need to be generated, which poses further challenges for explanations. To better measure explainability in CRS, we propose ten evaluation perspectives based on the concepts from conventional recommender systems together with the characteristics of CRS. We assess five existing CRS benchmark datasets using these metrics and observe the necessity of improving the explanation quality of CRS. To achieve this, we conduct manual and automatic approaches to extend these dialogues and construct a new CRS dataset, namely Explainable Recommendation Dialogues (E-ReDial). It includes 756 dialogues with over 2,000 high-quality rewritten explanations. We compare two baseline approaches to perform explanation generation based on E-ReDial. Experimental results suggest that models trained on E-ReDial can significantly improve explainability while introducing knowledge into the models can further improve the performance. GPT-3 in the in-context learning setting can generate more realistic and diverse movie descriptions. In contrast, T5 training on E-Redial can better generate clear reasons for recommendations based on user preferences. E-ReDial is available at https://github.com/Superbooming/E-ReDial.

Evaluation is crucial in the development process of task-oriented dialogue systems. As an evaluation method, user simulation allows us to tackle issues such as scalability and cost-efficiency, making it a viable choice for large-scale automatic evaluation. To help build a human-like user simulator that can measure the quality of a dialogue, we propose the following task: simulating user satisfaction for the evaluation of task-oriented dialogue systems. The purpose of the task is to increase the evaluation power of user simulations and to make the simulation more human-like. To overcome a lack of annotated data, we propose a user satisfaction annotation dataset, USS, that includes 6,800 dialogues sampled from multiple domains, spanning real-world e-commerce dialogues, task-oriented dialogues constructed through Wizard-of-Oz experiments, and movie recommendation dialogues. All user utterances in those dialogues, as well as the dialogues themselves, have been labeled based on a 5-level satisfaction scale. We also share three baseline methods for user satisfaction prediction and action prediction tasks. Experiments conducted on the USS dataset suggest that distributed representations outperform feature-based methods. A model based on hierarchical GRUs achieves the best performance in in-domain user satisfaction prediction, while a BERT-based model has better cross-domain generalization ability.

Today's open-domain conversational agents increase the informativeness of generated responses by leveraging external knowledge. Most of the existing approaches work only for scenarios with a massive amount of monolingual knowledge sources. For languages with limited availability of knowledge sources, it is not effective to use knowledge in the same language to generate informative responses. To address this problem, we propose the task of cross-lingual knowledge grounded conversation (CKGC), where we leverage large-scale knowledge sources in another language to generate informative responses. Two main challenges come with the task of cross-lingual knowledge grounded conversation: (1) knowledge selection and response generation in a cross-lingual setting; and (2) the lack of a test dataset for evaluation. To tackle the first challenge, we propose the curriculum self-knowledge distillation (CSKD) scheme, which utilizes a large-scale dialogue corpus in an auxiliary language to improve cross-lingual knowledge selection and knowledge expression in the target language via knowledge distillation. To tackle the second challenge, we collect a cross-lingual knowledge grounded conversation test dataset to facilitate relevant research in the future. Extensive experiments on the newly created dataset verify the effectiveness of our proposed curriculum self-knowledge distillation method for cross-lingual knowledge grounded conversation. In addition, we find that our proposed unsupervised method significantly outperforms the state-of-the-art baselines in cross-lingual knowledge selection.

Today's conversational agents often generate responses that not sufficiently informative. One way of making them more informative is through the use of of external knowledge sources with so-called Knowledge-Grounded Conversations (KGCs). In this paper, we target the Knowledge Selection (KS) task, a key ingredient in KGC, that is aimed at selecting the appropriate knowledge to be used in the next response. Existing approaches to Knowledge Selection (KS) based on learned representations of the conversation context, that is previous conversation turns, and use Maximum Likelihood Estimation (MLE) to optimize KS. Such approaches have two main limitations. First, they do not explicitly track what knowledge has been used in the conversation nor how topics have shifted during the conversation. Second, MLE often relies on a limited set of example conversations for training, from which it is hard to infer that facts retrieved from the knowledge source can be re-used in multiple conversation contexts, and vice versa.