Professional financial reports are the cornerstone of investment decision-making, demanding deep analytical reasoning and multimodal synthesis.While recent deep research systems excel in open-domain search tasks, they struggle with financial reporting, specifically in processing structured financial data, ensuring analytical depth, and integrating professional visualizations.To address this gap, we introduce FinSight (Financial InSight), the first multi-agent framework for end-to-end automation of professional multimodal financial reports.At its core, we propose a Code Agent with Variable Memory architecture, unifying financial data, domain tools, and agent modules into a programmable variable space to enable flexible data manipulation and reasoning via executable code.To guarantee report quality, FinSight adopts a Two-Stage Writing Framework with Generative Retrieval.It first distills raw data into structured Chain-of-Analysis segments, then synthesizes them into a coherent, citation-aware, and multimodal narrative complying with financial reporting norms.Additionally, an Iterative Vision-Enhanced Mechanism uses visual feedback to refine codegenerated charts to expert standards.Experiments on company and industry-level tasks demonstrate that FinSight significantly outperforms state-of-the-art deep research systems in factual accuracy, analytical depth, and presentation quality, validating the effectiveness of our framework.

Complex information needs in real-world search scenarios demand deep reasoning and knowledge synthesis across diverse sources, which traditional retrieval-augmented generation (RAG) pipelines struggle to address effectively. Current reasoning-based approaches face a key architectural challenge: they employ a single model to handle both high-level planning and detailed execution, resulting in inefficient reasoning and limited scalability. In this paper, we introduce HiRA, a hierarchical framework that separates strategic planning from specialized execution. Our approach decomposes complex search tasks into multiple subtasks, assigns each subtask to a domain-specific agent equipped with external tools and reasoning capabilities, and coordinates the results through a structured integration mechanism. This separation prevents execution details from disrupting high-level reasoning while enabling the system to leverage specialized expertise for different types of information processing. Experiments on four complex, cross-modal deep search benchmarks show that HiRA significantly outperforms state-of-the-art RAG and agent-based systems, highlighting the effectiveness of decoupled planning and execution for multi-step information seeking tasks. The code is available at https://github.com/RUC-NLPIR/HiRA.

Large Language Models (LLMs) have fundamentally transformed dense retrieval, upgrading backbones from discriminative encoders to generative architectures. However, a critical disconnect remains: while LLMs possess strong reasoning capabilities, current retrievers predominantly utilize them as static encoders, leaving their potential for complex reasoning unexplored. To address this, existing approaches typically adopt ''rewrite-then-retrieve'' pipelines to generate explicit Chain-of-Thought (CoT) rationales before retrieval. However, this incurs prohibitive latency. Conversely, implicit reasoning methods utilizing latent tokens offer efficiency but often suffer from semantic degeneration due to the lack of explicit supervision. In this paper, we propose LaSER, a novel self-distillation framework that internalizes explicit reasoning into the latent space of dense retrievers. Operating on a shared LLM backbone, LaSER introduces a dual-view training mechanism: an Explicit view that explicitly encodes ground-truth reasoning paths, and a Latent view that performs implicit latent thinking. To bridge the gap between these views, we design a multi-grained alignment strategy. Beyond standard output alignment, we introduce a trajectory alignment mechanism that synchronizes the intermediate latent states of the latent path with the semantic progression of the explicit reasoning segments. This allows the retriever to ''think'' silently and effectively without autoregressive text generation. Extensive experiments on both in-domain and out-of-domain reasoning-intensive benchmarks demonstrate that LaSER significantly outperforms state-of-the-art baselines. Furthermore, analyses across diverse backbones and model scales validate the robustness of our approach, confirming that our unified learning framework is essential for eliciting effective latent thinking. Our method successfully combines the reasoning depth of explicit CoT pipelines with the inference efficiency of standard dense retrievers. The code, model, and training data are available at https://github.com/RUC-NLPIR/LaSER.

Federated learning (FL) enables collaborative model training across multiple medical centers without sharing data, offering significant promise for privacy-preserving AI in healthcare. However, FL models often lack generalization across all participating clients (inside FL) and perform poorly when deployed to unseen clients (outside FL), particularly in heterogeneous domains. Current test-time adaptation methods for outside FL fail to address biases in personalized models toward source distributions, limiting their clinical applications. To tackle these challenges, we propose MSAFed, a generalized multi-stage adaptive FL framework that enhances both inside generalization and outside test-time adaptation. During pretraining, intra-client and inter-client contrastive learning with prototype-aware aggregation produces a generalized global model. An adaptive learning rate strategy further improves inside FL generalization. For unseen clients, source knowledge, including adaptive learning rates and prototypes, is leveraged to dynamically adapt the network architecture during test time. Experiments on three real-world multi-center medical datasets demonstrate the effectiveness of MSAFed, achieving superior performance on both inside and outside FL tasks.

Retrieval-augmented generation (RAG) has emerged as a pivotal technology in natural language processing, owing to its efficacy in generating factual content.However, its informative inputs and complex paradigms often lead to a greater variety of errors.Consequently, achieving automated on-policy assessment and error-oriented correction remains an unresolved issue.In this paper, we propose RAG-Critic, a novel framework that leverages a critic-guided agentic workflow to improve RAG capabilities autonomously.Specifically, we initially design a data-driven error mining pipeline to establish a hierarchical RAG error system.Based on this system, we progressively align an errorcritic model using a coarse-to-fine training objective, which automatically provides finegrained error feedback.Finally, we design a critic-guided agentic RAG workflow that customizes executor-based solution flows based on the error-critic model's feedback, facilitating an error-driven self-correction process.Experimental results across seven RAG-related datasets confirm the effectiveness of RAG-Critic, while qualitative analysis offers practical insights for achieving reliable RAG systems.Our dataset and code are available at https: //github.com/RUC-NLPIR/RAG-Critic.

Jiajie Jin, Xiaoxi Li, Guanting Dong, Yuyao Zhang, Yutao Zhu, Yongkang Wu, Zhonghua Li, Ye Qi, Zhicheng Dou. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Large language models (LLMs) exhibit remarkable generative capabilities but often suffer from hallucinations.Retrieval-augmented generation (RAG) offers an effective solution by incorporating external knowledge, but existing methods still face several limitations: additional deployment costs of separate retrievers, redundant input tokens from retrieved text chunks, and the lack of joint optimization of retrieval and generation.To address these issues, we propose RetroLLM, a unified framework that integrates retrieval and generation into a single, auto-regressive process, enabling LLMs to directly generate fine-grained evidence from the corpus with constrained decoding.Moreover, to mitigate false pruning in the process of constrained evidence generation, we introduce (1) hierarchical FM-Index constraints, which generate corpus-constrained clues to identify a subset of relevant documents before evidence generation, reducing irrelevant decoding space; and (2) a forward-looking constrained decoding strategy, which considers the relevance of future sequences to improve evidence accuracy.Extensive experiments on five open-domain QA datasets demonstrate RetroLLM's superior performance across both in-domain and out-ofdomain tasks.The code is available at https: //github.com/sunnynexus/RetroLLM.

Large reasoning models (LRMs) like OpenAI-o1 have demonstrated impressive long stepwise reasoning capabilities through large-scale reinforcement learning.However, their extended reasoning processes often suffer from knowledge insufficiency, leading to frequent uncertainties and potential errors.To address this limitation, we introduce Search-o1, a framework that enhances LRMs with an agentic retrievalaugmented generation (RAG) mechanism and a Reason-in-Documents module for refining retrieved documents.Search-o1 integrates an agentic search workflow into the reasoning process, enabling dynamic retrieval of external knowledge when LRMs encounter uncertain knowledge points.Additionally, due to the verbose nature of retrieved documents, we design a separate Reason-in-Documents module to deeply analyze the retrieved information before injecting it into the reasoning chain, minimizing noise and preserving coherent reasoning flow.Extensive experiments on complex reasoning tasks in science, mathematics, and coding, as well as six open-domain QA benchmarks, demonstrate the strong performance of Search-o1.This approach enhances the trustworthiness of LRMs in complex reasoning tasks, paving the way for advanced deep research systems.

Existing studies have optimized retrievalaugmented generation (RAG) across various sub-tasks, such as query understanding and retrieval refinement, but integrating these optimizations into a unified framework remains challenging.To tackle this problem, this work proposes RoleRAG, a unified RAG framework that achieves efficient multi-task processing through role-specific token optimization.RoleRAG comprises six modules, each handling a specific sub-task within the RAG process.Additionally, we introduce a query graph to represent the decomposition of the query, which can be dynamically resolved according to the decomposing state.All modules are driven by the same underlying LLM, distinguished by task-specific role tokens that are individually optimized.This design allows RoleRAG to dynamically activate different modules within a single LLM instance, thereby streamlining deployment and reducing resource consumption.Experimental results on five open-domain question-answering datasets demonstrate the effectiveness, generalizability, and flexibility of our framework.

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.

Large reasoning models (LRMs), such as OpenAI-o1 and DeepSeek-R1, demonstrate impressive long-horizon reasoning capabilities. However, their reliance on static internal knowledge limits their performance on complex, knowledge-intensive tasks and hinders their ability to produce comprehensive research reports requiring synthesis of diverse web information. To address this, we propose WebThinker, a deep research agent that empowers LRMs to autonomously search the web, navigate among web pages, and draft reports during the reasoning process. WebThinker integrates a Deep Web Explorer module, enabling LRMs to dynamically search, navigate, and extract information from the web when encountering knowledge gaps. It also employs an Autonomous Think-Search-and-Draft strategy, allowing the model to seamlessly interleave reasoning, information gathering, and report writing in real time. To further enhance research tool utilization, we introduce an RL-based training strategy via iterative online Direct Preference Optimization (DPO). Extensive experiments on complex reasoning benchmarks (GPQA, GAIA, WebWalkerQA, HLE) and scientific report generation tasks (Glaive) demonstrate that WebThinker significantly outperforms existing methods and strong proprietary systems. Our approach enhances LRM reliability and applicability in complex scenarios, paving the way for more capable and versatile deep research systems. The code is available at https://github.com/RUC-NLPIR/WebThinker.

Information Retrieval (IR) systems are crucial tools for users to access information, which have long been dominated by traditional methods relying on similarity matching. With the advancement of pre-trained language models, Generative Information Retrieval (GenIR) emerges as a novel paradigm, attracting increasing attention. Based on the form of information provided to users, current research in GenIR can be categorized into two aspects: (1) Generative Retrieval ( GR ) leverages the generative model’s parameters for memorizing documents, enabling retrieval by directly generating relevant document identifiers without explicit indexing. (2) Reliable Response Generation employs language models to directly generate information users seek, breaking the limitations of traditional IR in terms of document granularity and relevance matching while offering flexibility, efficiency, and creativity to meet practical needs. This article aims to systematically review the latest research progress in GenIR. We will summarize the advancements in GR regarding model training and structure, document identifier, incremental learning, and so on, as well as progress in reliable response generation in aspects of internal knowledge memorization, external knowledge augmentation, and so on. We also review the evaluation, challenges, and future developments in GenIR systems. This review aims to offer a comprehensive reference for researchers, encouraging further development in the GenIR field (Github Repository: https://github.com/RUC-NLPIR/GenIR-Survey ).

Retrieval-augmented generation have become central in natural language processing due to their efficacy in generating factual content. While traditional methods employ single-time retrieval, more recent approaches have shifted towards multi-time retrieval for multi-hop reasoning tasks. However, these strategies are bound by predefined reasoning steps, potentially leading to inaccuracies in response generation. This paper introduces MetaRAG, an approach that combines the retrieval-augmented generation process with metacognition. Drawing from cognitive psychology, metacognition allows an entity to self-reflect and critically evaluate its cognitive processes. By integrating this, MetaRAG enables the model to monitor, evaluate, and plan its response strategies, enhancing its introspective reasoning abilities. Through a three-step metacognitive regulation pipeline, the model can identify inadequacies in initial cognitive responses and fixes them. Empirical evaluations show that MetaRAG significantly outperforms existing methods.

Traditional search engines usually provide identical search results for all users, overlooking individual preferences. To counter this limitation, personalized search has been developed to re-rank results based on user preferences derived from query logs. Deep learning-based personalized search methods have shown promise, but they rely heavily on abundant training data, making them susceptible to data sparsity challenges. This paper proposes a Cognitive Personalized Search (CoPS) model, which integrates Large Language Models (LLMs) with a cognitive memory mechanism inspired by human cognition. CoPS employs LLMs to enhance user modeling and user search experience. The cognitive memory mechanism comprises sensory memory for quick sensory responses, working memory for sophisticated cognitive responses, and long-term memory for storing historical interactions. CoPS handles new queries using a three-step approach: identifying re-finding behaviors, constructing user profiles with relevant historical information, and ranking documents based on personalized query intent. Experiments show that CoPS outperforms baseline models in zero-shot scenarios.