Deep research is a retrieval-intensive task that requires iteratively retrieving evidence, reading across sources, and synthesizing source-grounded outputs. In practice, real-world deep research applications are of high workloads that require generating massive reports or conducting large-scale literature surveys. Such applications increasingly require batch processing capabilities, which are missing from traditional chat-oriented agents, limiting throughput when processing large volumes of structurally similar jobs. We introduce C3Flow (Concurrent Claude Code Workflow), a framework that transforms Claude Code from an interactive assistant into a SIMD-style (Single Instruction, Multiple Data) concurrent compute engine. C³Flow treats each agent instance as an isolated, schedulable unit capable of handling declarative multi-step tasks, multi-model routing, and comprehensive trajectory logging. On BrowseComp-zh, C³Flow improves pass@1 from 48.44% to 61.59% and pass@3 from 70.24% to 77.51% compared to standard function calling, while reducing average latency. For multi-hop fact verification, C3Flow achieves a 5.9 speedup over human annotators while maintaining 87.5% accuracy, demonstrating its effectiveness for production-scale deep research pipelines. Code is available at~ https://github.com/RAGenius/C3Flow.

Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of model safety during fine-tuning, we introduce LookAhead Tuning, a lightweight and effective data-driven approach that preserves safety during fine-tuning. The method introduces two simple strategies that modify training data by previewing partial answer prefixes, thereby minimizing perturbations to the model's initial token distributions and maintaining its built-in safety mechanisms. Comprehensive experiments demonstrate that LookAhead Tuning effectively maintains model safety without sacrificing robust performance on downstream tasks. Our findings position LookAhead Tuning https://github.com/zjunlp/LookAheadTuning as a reliable and efficient solution for the safe and effective adaptation of LLMs.

Retrieval Augmented Generation (RAG) has emerged as a promising solution to address hallucination issues in Large Language Models (LLMs). However, the integration of multiple retrieval sources, while potentially more informative, introduces new challenges that can paradoxically exacerbate hallucination problems. These challenges manifest primarily in two aspects: the sparse distribution of multi-source data that hinders the capture of logical relationships and the inherent inconsistencies among different sources that lead to information conflicts. To address these challenges, we propose MultiRAG, a novel framework designed to mitigate hallucination in multi-source retrieval-augmented generation through knowledge-guided approaches. Our framework introduces two key innovations: (1) a knowledge construction module that employs multi-source line graphs to efficiently aggregate logical relationships across different knowledge sources, effectively addressing the sparse data distribution issue; and (2) a sophisticated retrieval module that implements a multi-level confidence calculation mechanism, performing both graph-level and node-level assessments to identify and eliminate unreliable information nodes, thereby reducing hallucinations caused by inter-source inconsistencies. Extensive experiments on four multi-domain query datasets and two multi-hop QA datasets demonstrate that MultiRAG significantly enhances the reliability and efficiency of knowledge retrieval in complex multi-source scenarios. Our code is available in https://github.com/wuwenlong123/MultiRAG.

In real-world tasks like data management and Web search, index operations often exhibit strong skewness, unlike standard benchmarks with uniform data distribution. While learned indexes improve query and update efficiency, they typically fail to address the skewed workload access, often prioritizing a single performance metric at the cost of overall index effectiveness. Additionally, the full reliance on learned models can increase vulnerability to attacks, compromising system stability. To address these challenges, we propose ShapeShifter, an adaptive evolutionary structure based on traditional indexes, capable of dynamically adjusting node structures according to the workload. ShapeShifter introduces a node evolution strategy with workload-skew-aware policies to adaptively adjust and optimize the partial index structure, leveraging a hybrid mechanism that combines traditional and learned structures for robust performance and optimal time-space tradeoff under skewed workloads and extreme data conditions. The evaluation results show that ShapeShifter achieves the optimal tradeoff while maintaining robustness.

The first generation of learned indexes inherently achieved lower space overhead than traditional index structures, establishing this advantage as one of the pivotal research directions in index optimization. However, in their pursuit of peak performance, designers often significantly increase space overhead, which becomes infeasible in scenarios with limited storage space. Furthermore, the design of current learned indexes optimized for time-space tradeoff is flawed, as they collapse catastrophically under prevalent dense or duplicate insertion workloads. To address these challenges, we first quantitatively analyze the time-space correlation characteristics of learned indexes from a theoretical perspective and identify the core influencing factors. Based on this, time-space cost minimization function models are established and an updatable learned index framework, LIFT, is constructed. Furthermore, LIFT incorporates specifically designed structural adjustment mechanisms to effectively counter existing poisoning attacks, significantly enhancing index robustness without increasing time-space cost. Evaluation results demonstrate that LIFT consistently achieves the optimal time-space tradeoff across various workloads and datasets, outperforming all other state-of-the-art indexes.

Many distributed training techniques like Parameter Server and AllReduce have been proposed to take advantage of the increasingly large data and rich features. However, stragglers frequently occur in distributed training due to resource contention and hardware heterogeneity, which significantly hampers the training efficiency. Previous works only address part of the stragglers and could not adaptively solve various stragglers in practice. Additionally, it is challenging to use a systematic framework to address all stragglers because different stragglers require diverse data allocation and fault-tolerance mechanisms. Therefore, this paper proposes a unified distributed training framework called AntDT (Ant Distributed Training Framework) to adaptively solve the straggler problems. Firstly, the framework consists of four components, including the Stateful Dynamic Data Sharding service, Monitor, Controller, and Agent. These components work collaboratively to efficiently distribute workloads and provide a range of pre-defined straggler mitigation methods with fault tolerance, thereby hiding messy details of data allocation and fault handling. Secondly, the framework provides a high degree of flexibility, allowing for the customization of straggler mitigation solutions based on the specific circumstances of the cluster. Leveraging this flexibility, we introduce two straggler mitigation solutions, namely AntDT-ND for non-dedicated clusters and AntDT-DD for dedicated clusters, as practical examples to resolve various types of stragglers at Ant Group. Justified by our comprehensive experiments and industrial deployment statistics, AntDT outperforms other SOTA methods more than 3 × in terms of training efficiency. Additionally, in Alipay's homepage recommendation scenario, using AntDT reduces the training duration of the ranking model from 27.8 hours to just 5.4 hours.

The quest to identify local dense communities closely connected to predetermined seed nodes is vital across numerous applications. Given the seed nodes R, the R-subgraph density of a subgraph S is defined as traditional graph density of S with penalties on the nodes in S / R. The state-of-the-art (SOTA) anchored densest subgraph model, which is based on R-subgraph density, is designed to address the community search problem. However, it often struggles to efficiently uncover truly dense communities. To eliminate this issue, we propose a novel NR-subgraph density metric, a nuanced measure that identifies communities intimately linked to seed nodes and also exhibiting overall high graph density. We redefine the anchored densest subgraph search problem through the lens of NR-subgraph density and cast it as a Linear Programming (LP) problem. This allows us to transition into a dual problem, tapping into the efficiency and effectiveness of convex programming-based iterative algorithm. To solve this redefined problem, we propose two algorithms: FDP, an iterative method that swiftly attains near-optimal solutions, and FDPE, an exact approach that ensures full convergence. We perform extensive experiments on 12 real-world networks. The results show that our proposed algorithms not only outperform the SOTA methods by 3.6~14.1 times in terms of running time, but also produce subgraphs with superior internal quality.

Knowledge graphs (KGs) have become vitally important in modern recommender systems, effectively improving performance and interpretability. Fundamentally, recommender systems aim to identify user interests based on historical interactions and recommend suitable items. However, existing works overlook two key challenges: (1) an interest corresponds to a potentially large set of related items, and (2) the lack of explicit, fine-grained exploitation of KG information and interest connectivity. This leads to an inability to reflect distinctions between entities and interests when modeling them in a single way. Additionally, the granularity of concepts in the knowledge graphs used for recommendations tends to be coarse, failing to match the fine-grained nature of user interests. This homogenization limits the precise exploitation of knowledge graph data and interest connectivity. To address these limitations, we introduce a novel embedding-based model called InBox. Specifically, various knowledge graph entities and relations are embedded as points or boxes, while user interests are modeled as boxes encompassing interaction history. Representing interests as boxes enables containing collections of item points related to that interest. We further propose that an interest comprises diverse basic concepts, and box intersection naturally supports concept combination. Across three training steps, InBox significantly outperforms state-of-the-art methods like HAKG and KGIN on recommendation tasks. Further analysis provides meaningful insights into the variable value of different KG data for recommendations.

Based on the diversified application scenarios at Ant Group, we built the Ant Knowledge Graph Platform (AKGP). It has constructed numerous domain-specific knowledge graphs related to merchants, companies, accounts, products, and more. AKGP manages trillions of structured knowledge graphs, serving search, recommendation, risk control and other businesses. However, as the demand increasing for various workloads such as graph pattern matching, graph representation learning, and cross-domain knowledge reuse, the existing warehouse systems based on relational DBMS or graph databases are unable to meet the requirements. To address these issues, we propose KGFabric, an industrial-scale knowledge graph management system built on the distributed file system (DFS). KGFabric offers a nearline knowledge storage engine that utilizes a Semantic-enhanced Programmable Graph (SPG) model, which is compatible with the Labeled Property Graph (LPG) model. The data is persistently stored in DFS, such as HDFS, which leverages the POSIX file system API, making it suitable for deployment in multi-cloud environment at low cost. KGFabric provides a native graph-based and hybrid storage format that can serve as a shared backend for parallel graph computing systems, significantly accelerating the analysis of multi-workload. Additionally, KGFabric includes a graph fabric framework that minimizes data duplication and guarantees data security. KGFabric is able to manage Peta-scale data and has supported graph fabric and analysis with over 100 billion relations at Ant Group. We conduct experiments on various datasets to evaluate the performance of KGFabric. Compared with popular relational DBMS and graph databases, the storage space for semantic relations is reduced by over 90%. The performance of graph fabric improves by 21× in real-world workloads. In multi-hop semantic graph analysis, KGFabric enhances performance by 100×.