While Large Language Models (LLMs) augmented with search engines have achieved remarkable progress in open-domain question answering, their ability to adapt to a rapidly evolving world remains limited. A critical challenge lies in the need for complex temporal reasoning to answer real-world questions. Current Retrieval-Augmented Generation (RAG) methods primarily focus on retrieving the latest information but often fail to perform sophisticated temporal reasoning. To address this gap, we propose TimeRAG, a novel RAG framework designed to dynamically handle complex temporal reasoning tasks. TimeRAG operates through the iterative collaboration of two modules: (1) a temporal-semantic Query Decomposition (QD) module, which breaks down the original question into atomic time-event sub-questions to guide multi-step retrieval, and (2) a time-aware Answer Generation (AG) module, which analyzes temporal contexts, generates intermediate answers with confidence scores, and synthesizes the final answer upon reasoning completion. The system is trained in three stages: (1) time-aware supervised fine-tuning of the AG module, (2) imitation learning for the QD module to enhance temporal decomposition ability, and (3) reinforcement learning for end-to-end joint optimization to enhance temporal coherence across the entire system. Evaluations on three challenging benchmarks show that TimeRAG significantly outperforms existing methods, particularly on questions involving fast-changing real-world events and those grounded in false premises that require detection and correction of outdated or incorrect assumptions.

Active understanding and clarification of user intent is crucial for information-seeking systems based on Large Language Models (LLMs), as it enhances search efficiency and improves user experience for human-LLM interaction. While existing systems rely on domain-specific resources to generate clarifying questions, they face challenges when extended to open-domain scenarios due to the lack of human-LLM clarification data. In this paper, we propose ClariLM to synthesize large-scale clarification data and enhance the LLMs' clarification capability. Specifically, we design two key stages to prepare data: first, given a user question, the Clarification Facet Detection (CFD) stage employs a facet mining model learned from human-LLM conversation logs to predict realistic potential clarification candidates. Additionally, it incorporates direct predictions from powerful LLMs as supplements to guarantee comprehensive facet coverage. While CFD ensures high recall of facet candidates, the subsequent Optimal Facet Selection (OFS) stage synthesizes a set of new questions and employs a reasoning model to annotate the optimal facet for each question, which further improves the precision of ClariLM in clarification necessity prediction and optimal facet selection. The collected data are then applied for supervised fine-tuning, followed by constructing preference data for preference optimization. Experiments on our custom test set and two public benchmarks demonstrate that ClariLM significantly outperforms various baseline models across clarification necessity, clarifying question quality, and GPT-4-based comparative evaluation.

During interactions between users and Large Language Models (LLMs), users often engage in multi-turn questioning. Understanding the user's potential follow-up intents and generating follow-up question candidates for the user is crucial for enhancing their experience with LLMs. Existing methods for follow-up question generation mainly rely on hand-crafted rules, the internal knowledge of LLMs, or the integration of external knowledge. However, these approaches fail to effectively leverage real-world user follow-up intents when interacting with LLMs, resulting in generated questions that do not meet the needs of practical scenarios. In this paper, we propose FollowGPT, a model that mines user follow-up intents from user-LLM conversational logs. However, directly introducing raw conversation logs leads to significant noise and sparsity issues. Therefore, to address noises, FollowGPT adopts a hierarchical filtering strategy for data cleaning. To mitigate the sparsity issue, FollowGPT employs data synthesis methods to augment the log data across three dimensions: topic diversity, intent transition diversity, and negative sample diversity. The processed data is then consolidated into a new dataset named ShareFQG for both training and evaluation. Finally, we train FollowGPT using a two-stage training framework involving supervised fine-tuning and preference optimization. In our experiments, we evaluate on both the ShareFQG test set and a publicly available dataset, FollowupQG, using both automated metrics and GPT-4o-based comparative evaluation. The experimental results demonstrate that our method outperforms existing baselines in various metrics, including lexical similarity, semantic similarity, and GPT-4-based evaluation for follow-up question generation, demonstrating FollowGPT's effectiveness.

Search clarification is a critical user interface for open-domain conversational Web search, where generating high-quality facets for ambiguous or multi-facet queries significantly guides disambiguation and enhances the user's interaction experience. Recently, in-context learning with Large Language Models (LLMs) has emerged as a promising approach for facet generation by leveraging static or similarity-based demonstrations as prompts. However, existing methods predominantly rely on query similarity, failing to account for the multi-dimensional nature of query intents. This limitation can lead LLMs to generate incorrect or suboptimal facets misaligned with user needs. To address this challenge, we propose an intent-covering framework that improves clarification facet generation by selecting demonstrations that comprehensively cover the diverse intents underlying a given query. Specifically, we first train a generative model with beam search to predict potential intents and construct an intent-document graph to capture their semantic relationships. We then introduce a heuristic greedy algorithm that optimizes demonstration selection by maximizing intent coverage. Furthermore, since the order of demonstrations significantly affects generation quality, we develop a re-ranking model to optimize their sequence for better contextual alignment. Experiments demonstrate the superiority of our approach over strong baselines in various lexical and semantic evaluation metrics. Additionally, we conduct an in-depth analysis of how the number, order, and contextual relevance of demonstrations influence generation performance.

As a cornerstone of modern information access, search engines have become indispensable in everyday life. With the rapid advancements in AI and natural language processing (NLP) technologies, particularly large language models (LLMs), search engines have evolved to support more intuitive and intelligent interactions between users and systems. Conversational search, an emerging paradigm for next-generation search engines, leverages natural language dialogue to facilitate complex and precise information retrieval, thus attracting significant attention. Unlike traditional keyword-based search engines, conversational search systems enhance user experience by supporting intricate queries, maintaining context over multi-turn interactions, and providing robust information integration and processing capabilities. Key components such as query reformulation, search clarification, conversational retrieval, and response generation work in unison to enable these sophisticated interactions. In this survey, we explore the recent advancements and potential future directions in conversational search, examining the critical modules that constitute a conversational search system. We highlight the integration of LLMs in enhancing these systems and discuss the challenges and opportunities that lie ahead in this dynamic field. Additionally, we provide insights into real-world applications and robust evaluations of current conversational search systems, aiming to guide future research and development in conversational search.

Generating clarifying questions can effectively clarify users' complicated search intent in conversational search systems. However, existing methods based on pre-defined templates are inadequate in understanding explicit user intents, making generated questions monotonous or inaccurate in some cases. In this paper, we define the ''intent'' of a query as a verb representing the potential behavior, action, or task the user may take. We study generating clarifying questions from a new perspective by incorporating the intents explicitly to form ''intent-aware'' questions with high informativeness and accuracy. Since obtaining gold intent-aware questions is expensive, we propose a rule-based method and a continual learning model to generate intent-aware questions as weak supervision signals. The former leverages search results to mine contextual intent-aware words or phrases, and the latter relies on parallel corpora to paraphrase template-based questions by incorporating the intents. The generated weak supervision data are then applied to fine-tune a BART-based model for end-to-end intent-aware question generation. We also explore to prompt a large language model to generate intent-aware questions. Experimental results on a public clarification dataset demonstrate that our proposed methods improve users' search experience compared to existing methods.

Users' queries are usually vague, and their search intents tend to be ambiguous, thereby needing search clarification to clarify users' current intent by asking a clarifying question and providing several clickable sub-intent items as clarification options. However, in addition to drilling down the current query, users may also have exploratory needs that diverge from their current intent. For example, a user searching for the query "Cartier women watches'' may also potentially want to explore some parallel information by issuing queries such as "Rolex women watches'' or "Cartier women bracelets'', named exploratory queries in this paper. These exploratory needs are common during the search process yet cannot be satisfied by current search clarification approaches which typically stick to the sub-intents of the query. This paper focuses on mining exploratory queries as additional options to meet users' exploratory needs in conversational search systems. Specifically, we first design a rule-based model that generates exploratory queries based on the current query's top retrieved documents. Then, we propose using the data generated by the rule-based model to train a neural generation model through multi-task learning for further generalization. Finally, we borrow the in-context learning ability of the large language model to generate exploratory queries based on prompt engineering. We constructed an evaluation dataset based on human annotations and conduct an extensive set of experiments. The results show that our proposed methods generate higher-quality exploratory queries compared with several baselines.

Asking multi-turn clarifying questions has been applied in various conversational search systems to help recommend people, commodities, and images to users. However, its importance is still not emphasized in the Web search. In this paper, we make a step to extend the multi-turn clarification generation to Web search for clarifying users' ambiguous or faceted intents. Compared with other conversational search scenarios, Web search queries are more complicated, so clarification should be generated instead of being selected which is commonly applied in current studies. To this end, we first define the whole process of multi-turn Web search clarification composed of clarification candidate generation, optimal clarification selection, and document retrieval. Due to the lack of multi-turn open-domain clarification data, we first design a simple yet effective rule-based method to fit the above three components. After that, by utilizing the in-context learning and zero-shot instruction ability of large language models (LLMs), we implement clarification generation and selection by prompting LLMs with demonstrations and declarations, further improving the clarification effectiveness. To evaluate our proposed methods, we first measure whether our methods can improve the ability to retrieve documents. We also evaluate the quality of generated candidate facets. Experimental results show that, compared with existing single-turn methods for Web search clarification, our proposed framework is more suitable for open-domain Web search systems in asking multi-turn clarification questions to clarify users' ambiguous or faceted intents.

Search clarification in conversational search systems exhibits a clarification pane composed of several candidate aspect items and a clarifying question. To generate a pane, existing studies usually rely on unstructured document texts. However, important structured information in search results is not effectively considered, making the generated panes inaccurate in some cases. In this paper, we emphasize the importance of structured information in search results for improving search clarification. We propose enhancing unstructured documents with two kinds of structured information: one is "In-List'' relation obtained from HTML list structures, which helps extract groups of high-quality items with abundant parallel information. Another is "Is-A'' relation extracted from knowledge bases, which is helpful to generate good questions with explicit prompts. To avoid introducing excessive noises, we design a relation selection process to filter out ineffective relations. We further design a BART-based model for generating clarification panes. The experimental results show that the structured information is good supplement for generating high-quality clarification panes.

Supply Chain Platforms (SCPs) provide downstream industries with raw materials. Compared with traditional e-commerce platforms, data in SCPs is more sparse due to limited user interests. To tackle the data sparsity problem, one can apply Cross-Domain Recommendation (CDR) to improve the recommendation performance of the target domain with the source domain information. However, applying CDR to SCPs directly ignores hierarchical structures of commodities in SCPs, which reduce recommendation performance. In this paper, we take the catering platform as an example and propose GReS, a graphical CDR model. The model first constructs a tree-shaped graph to represent the hierarchy of different nodes of dishes and ingredients, and then applies our proposed Tree2vec method combining GCN and BERT models to embed the graph for recommendations. Experimental results show that GReS significantly outperforms state-of-the-art methods in CDR for SCPs.

Asking clarifying questions is an interactive way to effectively clarify user intent. When a user submits a query, the search engine will return a clarifying question with several clickable items of sub-intents for clarification. According to the existing definition, the key to asking high-quality questions is to generate good descriptions for submitted queries and provided items. However, existing methods mainly based on static knowledge bases are difficult to find descriptions for many queries because of the lack of entities within these queries and their corresponding items. For such a query, it is unable to generate an informative question. To alleviate this problem, we propose leveraging top search results of the query to help generate better descriptions because we deem that the top retrieved documents contain rich and relevant contexts of the query. Specifically, we first design a rule-based algorithm to extract description candidates from search results and rank them by various human-designed features. Then, we apply an learning-to-rank model and another generative model for generalization and further improve the quality of clarifying questions. Experimental results show that our proposed methods can generate more readable and informative questions compared with existing methods. The results prove that search results can be utilized to improve users' search experience for search clarification in conversational search systems.