Personalized large language models (LLMs) adapt model behavior to individual users to enhance user satisfaction, yet personalization can inadvertently distort factual reasoning. We show that when personalized LLMs face factual queries, there exists a phenomenon where the model generates answers aligned with a user’s prior history rather than the objective truth, resulting in personalization-induced hallucinations that degrade factual reliability and may propagate incorrect beliefs, due to representational entanglement between personalization and factual representations. To address this issue, we propose Factuality-Preserving Personalized Steering (FPPS), a lightweight inference-time approach that mitigates personalization-induced factual distortions while preserving personalized behavior. We further introduce PFQABench, the first benchmark designed to jointly evaluate factual and personalized question answering under personalization. Experiments across multiple LLM backbones and personalization methods show that FPPS substantially improves factual accuracy while maintaining personalized performance.

Tool-Integrated Reasoning (TIR) empowers large language models (LLMs) to tackle complex tasks by interleaving reasoning steps with external tool interactions. However, existing reinforcement learning methods typically rely on outcome- or trajectory-level rewards, assigning uniform advantages to all steps within a trajectory. This coarse-grained credit assignment fails to distinguish effective tool calls from redundant or erroneous ones, particularly in long-horizon multi-turn scenarios. To address this, we propose MatchTIR, a framework that introduces fine-grained supervision via bipartite matching-based turn-level reward assignment and dual-level advantage estimation. Specifically, we formulate credit assignment as a bipartite matching problem between predicted and ground-truth traces, utilizing two assignment strategies to derive dense turn-level rewards. Furthermore, to balance local step precision with global task success, we introduce a dual-level advantage estimation scheme that integrates turn-level and trajectory-level signals, assigning distinct advantage values to individual interaction turns. Extensive experiments on three benchmarks demonstrate the superiority of MatchTIR. Notably, our 4B model surpasses the majority of 8B competitors, particularly in long-horizon and multi-turn tasks. Our codes are available at https://anonymous.4open.science/r/MatchTIR.

In legal practice, judges apply the trichotomous dogmatics of criminal law, sequentially assessingthe elements of the offense, unlawfulness, and culpability to determine whether an individual’s conduct constitutes a crime. Although current legal large language models (LLMs) show promising accuracy in judgment prediction, they lack trichotomous reasoning capabilities due to the absence of an appropriate benchmark dataset, preventing them from predicting innocent outcomes. As a result, every input is automatically assigned a charge, limiting their practical utility in legal contexts. To bridge this gap, we introduce LJPIV, the first benchmark dataset for Legal Judgment Prediction with Innocent Verdicts. Adhering to the trichotomous dogmatics, we extend three widely-used legal datasets through LLM-based augmentation and manual verification. Our experiments with state-of-the-art legal LLMs and novel strategies that integrate trichotomous reasoning into zero-shot prompting and fine-tuning reveal: (1) current legal LLMs have significant room for improvement, with even the best models achieving an F1 score of less than 0.3 on LJPIV; and (2) our strategies notably enhance both in-domain and cross-domain judgment prediction accuracy, especially for cases resulting in an innocent verdict.

Legal mathematical reasoning is essential for applying large language models (LLMs) in high-stakes legal contexts, where outputs must be both mathematically accurate and procedurally compliant. However, existing legal LLMs lack structured numerical reasoning, and open-domain models, though capable of calculations, often overlook mandatory legal steps. To address this, we present LexNum, the first Chinese legal mathematical reasoning benchmark, covering three representative scenarios where each instance reflects legally grounded procedural flows. We further propose LexPam, a two-stage reinforcement learning framework for efficient legal reasoning training. Leveraging curriculum learning, we use a stronger teacher model to partition data into basic and challenging subsets. A lightweight 1.5B student model is then fine-tuned with Group Relative Policy Optimization, which avoids costly value networks and enables stable training from sparse, end-of-sequence rewards. The first stage improves accuracy and format; the second introduces a novel reward to guide procedural alignment via task-specific legal elements. Experiments show that existing models perform poorly on LexNum, while LexPam enhances both mathematical accuracy and legal coherence, and generalizes effectively across tasks and domains.

Query rewriting plays a pivotal role in Retrieval-Augmented Generation (RAG) by refining real-world queries of varying complexity. Existing approaches typically rely on outcome-supervised training or heuristic rules to guide the rewriting process. However, these paradigms often struggle to handle queries with varying levels of complexity, posing over- and under-refinement problems. We identify the root cause of these issues as the absence of supervision signals for intermediate steps. To fully construct and utilize such signals, we propose Q-PRM, a novel query rewriting framework. Q-PRM reformulates the rewriting process as a Markov Decision Process (MDP) composed of atomic rewriting steps. In this way, Q-PRM can apply process-level supervision to each atomic step according to the query type, offering more targeted and effective guidance. Q-PRM comprises three key stages: (1) applying Monte Carlo Tree Search to generate step-level process supervision signals; (2) performing reinforced self-training for progressive process refinement; and (3) employing PRM-guided decoding during inference. Experiments on several open-domain QA benchmarks demonstrate that Q-PRM consistently outperforms baselines across different levels of query complexity.

Retrieval-augmented generation (RAG) has proven effective in enhancing the knowledge coverage of large language models (LLMs) and mitigating hallucinations by incorporating external retrieved documents. However, documents deemed relevant by the retriever are not necessarily helpful for answer generation, and including misleading information can even degrade performance. Existing efforts to estimate document utility often rely on the downstream generation performance, which conflates the influence of external documents with the intrinsic knowledge of the LLM, thereby obscuring the actual contribution of the retrieved content. To address this, this paper proposes Uplit-RAG, a uplift-driven knowledge preference alignment framework for RAG. Specifically, we first propose an uplift-based definition of document utility that quantifies each document’s marginal benefit over the LLM’s internal knowledge. We then optimize the reranker with three alignment objectives to identify and prioritize documents based on their uplift. This enables dynamic selection of documents that address the LLM’s knowledge gaps, going beyond fixed top-k selection, while reducing reference redundancy and the computational overhead of the LLM’s input. Extensive experiments demonstrate the effectiveness of Uplift-RAG.

In this paper, we propose CitaLaw, the first benchmark designed to evaluate LLMs’ ability to produce legally sound responses with appropriate citations. CitaLaw features a diverse set of legal questions for both laypersons and practitioners, paired with a comprehensive corpus of law articles and precedent cases as a reference pool. This framework enables LLM-based systems to retrieve supporting citations from the reference corpus and align these citations with the corresponding sentences in their responses. Moreover, we introduce syllogism-inspired evaluation methods to assess the legal alignment between retrieved references and LLM-generated responses, as well as their consistency with user questions. Extensive experiments on 2 open-domain and 7 legal-specific LLMs demonstrate that integrating legal references substantially enhances response quality. Furthermore, our proposed syllogism-based evaluation method exhibits strong agreement with human judgments.

Unlike traditional search engines that present ranked lists of webpages, generative search engines rely solely on in-line citations as the key gateway to original real-world webpages, making it crucial to examine whether LLM-generated citations have biases—particularly for politically sensitive queries. To investigate this, we first construct AllSides-2024, a new dataset comprising the latest real-world news articles (Jan. 2024 - Dec. 2024) labeled with left- or right-leaning stances. Through systematic evaluations, we find that LLMs exhibit a consistent tendency to cite left-leaning sources at notably higher rates compared to traditional retrieval systems (e.g., BM25 and dense retrievers). Controlled experiments further reveal that this bias arises from a preference for media outlets identified as left-leaning, rather than for left-oriented content itself. Meanwhile, our findings show that while LLMs struggle to infer political bias from news content alone, they can almost perfectly recognize the political orientation of media outlets based on their names. These insights highlight the risk that, in the era of generative search engines, information exposure may be disproportionately shaped by specific media outlets, potentially shaping public perception and decision-making.

Personalized search systems in e-commerce platforms increasingly involve user interactions with AI assistants, where users consult about products, usage scenarios, and more. Leveraging consultation to personalize search services is trending. Existing methods typically rely on semantic similarity to align historical consultations with current queries due to the absence of ‘value’ labels, but we observe that semantic similarity alone often fails to capture the true value of consultation for personalization. To address this, we propose a consultation value assessment framework that evaluates historical consultations from three novel perspectives: (1) Scenario Scope Value, (2) Posterior Action Value, and (3) Time Decay Value. Based on this, we introduce VAPS, a value-aware personalized search model that selectively incorporates high-value consultations through a consultation–user action interaction module and an explicit objective that aligns consultations with user actions. Experiments on both public and commercial datasets show that VAPS consistently outperforms baselines in both retrieval and ranking tasks. Codes are available at https://github.com/E-qin/VAPS.

Continual pre-training (CPT) has been an important approach for adapting language models to specific domains or tasks. In this paper, we comprehensively study its key designs to balance the new abilities while retaining the original abilities, and present an effective CPT method that can greatly improve the Chinese language ability and scientific reasoning ability of LLMs. To achieve it, we design specific data mixture and curriculum strategies based on existing datasets and synthetic high-quality data. Concretely, we synthesize multidisciplinary scientific QA pairs based on related web pages to guarantee the data quality, and also devise the performance tracking and data mixture adjustment strategy to ensure the training stability. For the detailed designs, we conduct preliminary studies on a relatively small model, and summarize the findings to help optimize our CPT method. Extensive experiments on a number of evaluation benchmarks show that our approach can largely improve the performance of Llama-3 (8B), including both the general abilities (+8.81 on C-Eval and +6.31 on CMMLU) and the scientific reasoning abilities (+12.00 on MATH and +4.13 on SciEval). Our model, data, and codes are available at https://github.com/RUC-GSAI/Llama-3-SynE.

Personalized product search aims to retrieve and rank items that match users’ preferences and search intent. Despite their effectiveness, existing approaches typically assume that users’ query fully captures their real motivation. However, our analysis of a real-world e-commerce platform reveals that users often engage in relevant consultations before searching, indicating they refine intents through consultations based on motivation and need. The implied motivation in consultations is a key enhancing factor for personalized search. This unexplored area comes with new challenges including aligning contextual motivations with concise queries, bridging the category-text gap, and filtering noise within sequence history. To address these, we propose a Motivation-Aware Personalized Search (MAPS) method. It embeds queries and consultations into a unified semantic space via LLMs, utilizes a Mixture of Attention Experts (MoAE) to prioritize critical semantics, and introduces dual alignment: (1) contrastive learning aligns consultations, reviews, and product features; (2) bidirectional attention integrates motivation-aware embeddings with user preferences. Extensive experiments on real and synthetic data show MAPS outperforms existing methods in both retrieval and ranking tasks. Code and supplementary materials are available at: https://github.com/E-qin/MAPS.

Text embeddings from PLM-based models enable a wide range of applications, yet their performance often degrades on longer texts. In this paper, we introduce a phenomenon we call Length Collapse, where embeddings of longer texts tend to cluster together. This clustering results in a distributional inconsistency between the embeddings of short and long texts. We further investigate how these differences contribute to the performance decline observed with longer texts across various downstream tasks. Through a rigorous theoretical analysis of the self-attention mechanism, which acts as a low-pass filter in PLM-based models, we demonstrate that as text length increases, the strength of low-pass filtering intensifies, causing embeddings to retain more low-frequency components. As a result, input token features become more similar, leading to clustering and ultimately the collapse of embeddings for longer texts. To address this issue, we propose a simple method, TempScale, which mitigates the Length Collapse phenomenon. By narrowing the gap in low-pass filtering rates between long and short texts, TempScale ensures more consistent embeddings across different text lengths. This approach leads to performance improvements of 0.94% on MTEB and 1.10% on LongEmbed, which focuses specifically on long-context retrieval, providing strong evidence for the validity of our analysis. The source code is available at bluehttps://github.com/Yuqi-Zhou/Length_Collapse.