Named entity recognition (NER) extracts entity spans and assigns them to predefined types. Large language models (LLMs) are increasingly used for NER, but most LLM-based methods still rely on a mainly semantic view, judging whether a candidate span is compatible with an entity type. Although effective, this view can miss entities with weak type semantics, over-accept semantically salient modifiers, and leave little chance to recover spans omitted by early candidate extraction. We argue that entity spans are also traceable in sentence form: local structure and prepositions provide cues about a span’s position and function. We propose SSDA-NER, a Semantic-Structural Dual-path Arbitration framework that combines semantic understanding with structural analysis. Specifically, the Semantic path checks candidate type compatibility, while the Structure path withholds explicit semantic guidance during its first stage and identifies possible mentions through structural analysis. The Arbitration Agent then compares their rationales and confidence scores to adjudicate disagreements. Experiments on nine datasets with three backbone LLMs show that the two views are complementary: the Structure path substantially supplements entities missed by the Semantic path on some datasets, and arbitration outperforms existing multi-agent methods in most evaluated settings. Source code is available at https://github.com/MrHoux/SSDA-NER.

AI-generated text detection is commonly framed as a binary document-level judgment about whether a text is human-written or machine-generated. This framing breaks down for mixed-origin writing, where content origin and expression origin may differ. We cast mixed-origin detection as dimension-to-composition source attribution, inferring content origin and expression origin before composing them into four collaboration types. We propose *Dimension-to-Composition Routing* (**D2C-Routing**), which routes content-side and expression-side evidence to supervised dimension heads before a learned gated composition layer predicts the final label. On MixD2C, a reconstructed split derived from the HART mixed-origin benchmark, our disclosed D2C-Routing-based detector system reaches 0.8603 four-way Avg TPR@1%FPR, 6.5 points above the same-split RACE-local rerun. Core ablations support the routing design, while error analysis shows that distinguishing AI-content/human-expression from fully AI-generated text remains the hardest boundary. Code is available at https://github.com/bystander563/d2c-routing-artifact.

Large reasoning models improve problem-solving performance by generating extended Chain-of-Thought, but often suffer from overthinking, producing unnecessarily long and redundant reasoning traces. Prior works mainly focus on directly compressing reasoning length, or selecting reasoning modes globally based on task difficulty, without accounting for the dynamic and heterogeneous nature of the reasoning process itself. This lack of consideration for inference-time diversity limits their ability to adapt reasoning strategies flexibly. To address this gap, we propose Interleaved System-1/2 Reasoning (InterSR), a cognitively inspired framework that dynamically alternates between System 1 (fast) and System 2 (slow) reasoning modes during generation. Central to InterSR is a learnable system switching signal that reflects the necessity for deeper reasoning at each intermediate step and enables fine-grained control over cognitive mode transitions. Experiments across mathematical and cross-domain benchmarks show that InterSR effectively captures internal reasoning signals to reduce token usage while maintaining strong and robust accuracy. We release our code at https://github.com/buaadh/InterSR.

Code retrieval is becoming central to coding agents, but agentic coding requires more than matching a natural-language query to an isolated snippet. Given a user request, a coding agent needs to navigate a concrete repository state, locate relevant files and functions, gather supporting context, and filter similar in-repository distractors. Existing code retrieval benchmarks mainly evaluate docstring-to-function or snippet-level matching, thereby missing this requirement-driven repository search problem. To address this gap, we introduce CORE-Bench, a comprehensive benchmark for code retrieval in the era of agentic coding. CORE-Bench evaluates code retrieval ability at three levels: code understanding, issue-to-edit localization, and broader context retrieval. Built from curated code-search tasks and SWE-bench-series instances, CORE-Bench contains over 180K queries and 106K broader-context relevance labels. Experiments with representative embedding models show a sharp drop from traditional code search to code retrieval in agentic coding settings. Simple supervised fine-tuning of existing embedding models significantly improves performance in this setting, suggesting substantial room for further progress.

Rubric generation improves the interpretability and actionability of large language model evaluation, but in online-to-offline (O2O) local-life AI search, user queries often reflect more complex needs, such as consumption intent, location constraints, and scenario preferences. Directly generating rubrics from a single query tends to produce generic criteria that fail to satisfy AI search evaluation needs, while leveraging candidate responses, preference samples, or external evidence at scale remains challenging. To address these limitations, we propose Category-Aware Rubric Generation (CaRG), a category-aware evaluation dimension candidate pool framework. Through embedding-based clustering, representative selection, and iterative LLM-based consolidation, CaRG constructs reusable candidate pools of both general and category-specific evaluation dimensions from representative historical queries. Subsequently, these candidate pools are leveraged to generate query-level rubrics with concrete positive and negative scoring criteria for target queries. Experimental results demonstrate that CaRG consistently outperforms Direct-Rubric and shows more stable overall quality than existing query-only baselines, while achieving the strongest overall agreement with production golden rubric rankings. Our repository is available at https://github.com/PlussLe/CaRG.

Awakening dormant users, who remain engaged but exhibit low conversion, is a key driver for incremental GMV growth in e-commerce platforms. Existing methods often underperform since they typically rely on single-step estimation of an item’s intrinsic value (e.g., immediate click probability). This overlooks the instrumental effect of items, where specific interactions act as triggers to shape latent intent and drive subsequent decisions along a conversion trajectory. To bridge this gap, we propose RoleGen, a framework that synergizes a Conversion Trajectory Reasoner with a Generative Collaborative Executor. Specifically, the LLM-based Reasoner explicitly models the context-dependent Functional Role of items to reconstruct intent evolution. It further uses counterfactual role exploration to mitigate interest collapse and generate diverse conversion-oriented candidates. These reasoned candidate items are then injected into the Executor, which is optimized via a feedback-driven co-training mechanism to ensure grounded execution. Extensive offline experiments and online A/B testing on a large-scale e-commerce platform show a 5.82% gain in Recall@1 and a 6.71% increase in online order volume, confirming RoleGen’s effectiveness in activating dormant users.

Generative Retrieval (GR) introduces a new information retrieval paradigm that directly generates unique document identifiers (DocIDs). The key challenge of GR lies in creating effective yet discrete DocIDs that preserve semantic relevance for similar documents while differentiating dissimilar ones. However, existing methods generate DocIDs solely based on the textual content of documents, which may result in DocIDs with weak semantic connections for similar documents due to variations in expression. Therefore, we propose using queries as a bridge to connect documents with varying relevance levels for learning improved DocIDs. In this paper, we propose Multi-lEvel Relevance document identifier learning for Generative rEtrieval (MERGE), a novel approach that utilizes multi-level document relevance to learn high-quality DocIDs. MERGE incorporates three modules: a multi-relevance query-document alignment module to effectively align document representations with related queries, an outer-level contrastive learning module to capture binary-level relevance, and an inner-level multi-level relevance learning module to distinguish documents with different relevance levels. Our approach encodes rich hierarchical semantic information and maintains uniqueness across documents. Experimental results on real-world multilingual e-commerce search datasets demonstrate that MERGE significantly outperforms existing methods, underscoring its effectiveness. The source code is available at https://github.com/zhangfw123/MERGE.

Food delivery search aims to quickly retrieve deliverable items that meet users’ needs, typically requiring faster and more accurate query understanding compared to traditional e-commerce search. Generative retrieval (GR), an emerging search paradigm, harnesses the advanced query understanding capabilities of large language models (LLMs) to enhance the retrieval of results for complex and long-tail queries in food delivery search scenarios. However, there are still challenges in deploying GR to online scenarios: 1) the large scale of items; 2) latency constraints unmet by LLM inference in online retrieval; and 3) strong location-based service restrictions on generated items. To explore the application of GR in food delivery search, we optimize both offline training and online deployment, proposing Hierarchical semantic representation enhancement for Generative Retrieval (HierGR). Specifically, for the generation of semantic IDs, we propose an optimization method that refines the residual quantization process to generate hierarchically semantic IDs for items. Additionally, to successfully deploy on a well-known food delivery platform, we utilize the query cache mechanism and integrate the GR model with the online dense retrieval model to fulfill real-world search requirements. Online A/B testing results show that our proposed method increases the number of online orders by 0.68% for complex search intents. The source code is available at https://github.com/zhangfw123/HierGR.

Large Language Models (LLMs) have recently gained significant interest due to their impressive results in various natural language tasks. However, their application to sentence embeddings is still under active research. In this work, we introduce PromptEOL, a simple and efficient method designed to enhance LLM performance on sentence embeddings with a one-word limitation. We further integrate PromptEOL with in-context learning and alignment to leverage LLMs in two settings: without fine-tuning and with fine-tuning. Our extensive experiments show that PromptEOL enables LLMs to generate superior sentence embeddings without fine-tuning, outperforming contrastive learning methods. Additionally, with fine-tuning, a 2.7B parameter model using PromptEOL surpasses the performance of a 4.8B parameter model from previous methods. We also analyze how scaling model parameters, from 125 million to 66 billion, impacts sentence embedding performance.

To overcome the overparameterized problem in Pre-trained Language Models (PLMs), pruning is widely used as a simple and straightforward compression method by directly removing unimportant weights. Previous first-order methods successfully compress PLMs to extremely high sparsity with little performance drop. These methods, such as movement pruning, use first-order information to prune PLMs while fine-tuning the remaining weights. In this work, we argue fine-tuning is redundant for first-order pruning, since first-order pruning is sufficient to converge PLMs to downstream tasks without fine-tuning. Under this motivation, we propose Static Model Pruning (SMP), which only uses first-order pruning to adapt PLMs to downstream tasks while achieving the target sparsity level. In addition, we also design a new masking function and training objective to further improve SMP. Extensive experiments at various sparsity levels show SMP has significant improvements over first-order and zero-order methods. Unlike previous first-order methods, SMP is also applicable to low sparsity and outperforms zero-order methods. Meanwhile, SMP is more parameter efficient than other methods due to it does not require fine-tuning.

We propose PromptBERT, a novel contrastive learning method for learning better sentence representation. We firstly analysis the drawback of current sentence embedding from original BERT and find that it is mainly due to the static token embedding bias and ineffective BERT layers. Then we propose the first prompt-based sentence embeddings method and discuss two prompt representing methods and three prompt searching methods to make BERT achieve better sentence embeddings .Moreover, we propose a novel unsupervised training objective by the technology of template denoising, which substantially shortens the performance gap between the supervised and unsupervised settings. Extensive experiments show the effectiveness of our method. Compared to SimCSE, PromptBert achieves 2.29 and 2.58 points of improvement based on BERT and RoBERTa in the unsupervised setting.

Hierarchical text classification aims to leverage label hierarchy in multi-label text classification. Existing methods encode label hierarchy in a global view, where label hierarchy is treated as the static hierarchical structure containing all labels. Since global hierarchy is static and irrelevant to text samples, it makes these methods hard to exploit hierarchical information. Contrary to global hierarchy, local hierarchy as a structured labels hierarchy corresponding to each text sample. It is dynamic and relevant to text samples, which is ignored in previous methods. To exploit global and local hierarchies, we propose Hierarchy-guided BERT with Global and Local hierarchies (HBGL), which utilizes the large-scale parameters and prior language knowledge of BERT to model both global and local hierarchies. Moreover, HBGL avoids the intentional fusion of semantic and hierarchical modules by directly modeling semantic and hierarchical information with BERT. Compared with the state-of-the-art method HGCLR, our method achieves significant improvement on three benchmark datasets.

Recently, some span-based methods have achieved encouraging performances for joint aspect-sentiment analysis, which first extract aspects (aspect extraction) by detecting aspect boundaries and then classify the span-level sentiments (sentiment classification). However, most existing approaches either sequentially extract task-specific features, leading to insufficient feature interactions, or they encode aspect features and sentiment features in a parallel manner, implying that feature representation in each task is largely independent of each other except for input sharing. Both of them ignore the internal correlations between the aspect extraction and sentiment classification. To solve this problem, we novelly propose a hierarchical interactive network (HI-ASA) to model two-way interactions between two tasks appropriately, where the hierarchical interactions involve two steps: shallow-level interaction and deep-level interaction. First, we utilize cross-stitch mechanism to combine the different task-specific features selectively as the input to ensure proper two-way interactions. Second, the mutual information technique is applied to mutually constrain learning between two tasks in the output layer, thus the aspect input and the sentiment input are capable of encoding features of the other task via backpropagation. Extensive experiments on three real-world datasets demonstrate HI-ASA’s superiority over baselines.

The rapid development of knowledge graphs (KGs), such as Freebase and WordNet, has changed the paradigm for AI-related applications. However, even though these KGs are impressively large, most of them are suffering from incompleteness, which leads to performance degradation of AI applications. Most existing researches are focusing on knowledge graph embedding (KGE) models. Nevertheless, those models simply embed entities and relations into latent vectors without leveraging the rich information from the relation structure. Indeed, relations in KGs conform to a three-layer hierarchical relation structure (HRS), i.e., semantically similar relations can make up relation clusters and some relations can be further split into several fine-grained sub-relations. Relation clusters, relations and sub-relations can fit in the top, the middle and the bottom layer of three-layer HRS respectively. To this end, in this paper, we extend existing KGE models TransE, TransH and DistMult, to learn knowledge representations by leveraging the information from the HRS. Particularly, our approach is capable to extend other KGE models. Finally, the experiment results clearly validate the effectiveness of the proposed approach against baselines.