论文

基于实证对 "Enhanced" 与 "Agentic" RAG 范式进行评估，为真实场景中选取最有效的 RAG 设计（兼顾性能与成本）提供指导。An empirically driven evaluation of the "Enhanced" and "Agentic" RAG paradigms is conducted, offering guidance on selecting the most effective RAG design for real-world applications, considering both performance and costs.

本文提出了 SeedRG，一个用于缓解 knowledge leakage 并应对 benchmark aging 问题的半合成 benchmark 生成 pipeline。SeedRG is introduced, a semi-synthetic benchmark generation pipeline that mitigates knowledge leakage and addresses the issue of benchmark aging.

本文提出一个无偏的多进程 evaluation 框架，能够有效分散 client 端负载，从而在每秒数千次 query 以上的生产规模下实现对 LLM 的精确、可复现 profiling。This work proposes an unbiased, multi-process evaluation framework that effectively distributes client-side load, enabling accurate, reproducible profiling of LLMs at production scales exceeding thousands of queries per second.

本文提出 RAGCap-Bench，一个面向能力的 benchmark，用于对 agentic RAG workflow 中的中间任务进行细粒度评测，并构建了典型 LLM 错误的分类体系以设计针对性评测问题。This work proposes RAGCap-Bench, a capability-oriented benchmark for fine-grained evaluation of intermediate tasks in agentic RAG workflows, and constructs a taxonomy of typical LLM errors to design targeted evaluation questions.

本文提出 AgenticRAGTracer，这是首个主要由大语言模型自动构建、专为支持逐步验证而设计的 Agentic RAG 基准。AgenticRAGTracer is introduced, the first Agentic RAG benchmark that is primarily constructed automatically by large language models and designed to support step-by-step validation, and is primarily constructed automatically by large language models and designed to support step-by-step validation.

本工作提出 AlphaEval，一个基于真实生产环境的基准，包含来自七家在其核心业务中部署 AI Agent 的公司的 94 个任务，覆盖六个 O*NET (Occupational Information Network) 领域；并贡献了一套从需求到基准的构建框架，将从需求到评估的完整流程标准化。This work presents AlphaEval, a production-grounded benchmark of 94 tasks sourced from seven companies deploying AI agents in their core business, spanning six O*NET (Occupational Information Network) domains, and contributes a requirement-to-benchmark construction framework that standardizes the entire pipeline from requirement to evaluation.

本文提出 DashboardQA，这是首个明确设计用于评估视觉-语言 GUI Agent 对真实世界仪表板理解与交互能力的基准，结果表明交互式仪表板推理对所有受评估的 VLM 而言都是一项具有挑战性的任务。DashboardQA is introduced, the first benchmark explicitly designed to assess how vision-language GUI agents comprehend and interact with real-world dashboards, and indicates that interactive dashboard reasoning is a challenging task overall for all the VLMs evaluated.

提出 RoboDojo，一个面向通用机器人操作策略综合评估的仿真-真机统一基准，将 30 种策略集成到 XPolicyLab 并在 RoboDojo 上进行评测，建立了公开的排行榜与系统性的策略性能分析。RoboDojo is introduced, a unified sim-and-real benchmark for comprehensive evaluation of generalist robot manipulation policies that integrates 30 policies into XPolicyLab and evaluates them on RoboDojo, establishing a public leaderboard and systematic analysis of current policy performance.

确认了丰富资源语言与代表性不足语言之间在数学推理性能上存在持续差距，性能更优主要与更强的指令遵循能力相关，并提出了完全开源的数据集、数据采集流程与评估框架。A persistent gap in mathematical reasoning performance between high-resource and underrepresented languages is confirmed, with stronger results largely associated with better instruction-following ability, and a fully open-source dataset, data acquisition pipeline, and evaluation framework is introduced.

该工作提出 HETERQA，一个包含 857 个 QA 对的综合性基准，涵盖五个异构来源的记录检索，并表明 HETERQA 为异构来源下的记录检索提供了有效的测试平台，为未来检索方法留下了显著空间。This work introduces HETERQA, a comprehensive benchmark with 857 QA pairs for record retrieval over five heterogeneous sources and indicates that HETERQA provides an effective testbed for record retrieval over heterogeneous sources and leaves substantial room for future retrieval methods.

将预测的味觉空间作为基于内容的检索索引，对 309 项条目池的排序比 CLAP-text 基线（处于随机水平）忠实得多；ridge probes 与 audio-bandstop knockout 在已记载的声-味对应关系上读出了最强表征。Operationalised as a content-based retrieval index, the predicted taste space ranks a 309-item pool far more faithfully than a CLAP-text baseline, which sits at chance; ridge probes and an audio-bandstop knockout read the strongest representations against documented sound-taste correspondences.

数据混合（而非过滤）是构建高质量训练数据集的关键：以指令型数据为主的混合在扩展时优于以描述型数据为主的混合，且规模越大优势越明显。It is found that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales.

静脉识别是一种安全生物特征技术，常受限于标注数据稀缺与成像差异；而面向自然图像设计的增强策略可能破坏其关键的细粒度拓扑与纹理。本文提出 AGVBench，在 5 个公开掌/指静脉数据集、7 种骨干网络（含经典 CNN、视觉 Transformer 及静脉专用模型）上评测 30 种代表性增强策略。结果显示，多图混合类方法（如 MixUp、PuzzleMix……Vein recognition is a secure biometric technology often constrained by limited annotated data and imaging variations. While data augmentation mitigates this, strategies designed for natural images may disrupt the fine-grained topology and textures essential for identity discrimination. We present AGVBench, which evaluates 30 representative augmentation strategies on five public palm- and finger-vein datasets with seven backbone architectures, covering classic CNNs, vision transformers, and vein-specific recognition models. Our results show that multi-image mixing methods (e.g., MixUp, PuzzleMi

提出"自主策略演化"评估范式：在固定交互预算下，由 harness-model Agent 反复编辑可执行策略系统；并在 EvoPolicyGym 中实例化，该基准基于一组紧凑型交互式 RL 环境构建，用于评测 Agent 如何迭代改进已探索策略。This work introduces Autonomous Policy Evolution, a controlled evaluation setting in which a harness-model agent repeatedly edits an executable policy system under a fixed interaction budget, and instantiates this setting in EvoPolicyGym, a benchmark built from compact interactive RL environments that evaluates how agents iteratively improve explored policies.

本文适配了一款专家混合扩散语言模型 DiffusionGemma-26B，并在医学视觉问答数据集上，使用相同的 LoRA 配置将其与同规模的自回归模型 Gemma-4-26B 进行基准对比，由对冗长度鲁棒的 LLM 裁判打分。This work adapts a mixture-of-experts diffusion language model, DiffusionGemma-26B, and benchmark it against its same-size AR sibling Gemma-4-26B under an identical LoRA recipe on medical visual question answering datasets, scored by a verbosity-robust LLM judge.

在三种多模态 LLM 主干模型上，MRPO 均稳定优于标准 GRPO 及一项最新的 RL 基线；在 Qwen3-VL-8B-Instruct 上甚至超越规模显著更大的医学 MLLM（如 HuatuoGPT-Vision-34B）2.79 分。Across three multimodal LLM backbones, MRPO consistently outperforms standard GRPO and a recent RL baseline, and on Qwen3-VL-8B-Instruct even surpasses substantially larger medical MLLMs such as HuatuoGPT-Vision-34B by 2.79 points.

BeyondArena 是首个面向表格数据的统一整体基准，支持多种任务类型（IID、时序、分组），覆盖样本量与特征维度的不同尺度，并涵盖来自广泛学科的多样化特征类型。BeyondArena is the first unified holistic benchmark for tabular data that supports diverse task types (IID, temporal, grouped), across sample size and feature dimensionality scales, with diverse feature types from a broad range of disciplines.

提出 PerceptionRubrics——一个基于评分量表的评估框架，旨在弥合饱和的基准分数与真实场景脆弱性之间的差距，并验证了严格的感知保真是可靠生成的前提。PerceptionRubrics is introduced, a rubric-based evaluation framework that addresses the gap between saturated benchmark scores and real-world brittleness, validating that strict perceptual fidelity is the prerequisite for reliable generation.

提出 MIRROR——一个统一的跨表层框架，在显式新颖性约束下以检索到的上下文为条件生成候选，并执行记忆引导的蒙特卡洛树搜索，使检索可影响搜索先验，同时避免提示词级别的复制。MIRROR is presented, a unified cross-surface framework that performs memory-guided Monte Carlo tree search while conditioning candidate generation on retrieved context under an explicit novelty constraint, allowing retrieval to inform search priors without enabling prompt copying.

本工作将带外防御组织为经典完整性保护、引用监控与最小权限的具体实例，对它们覆盖与未覆盖的内容进行结构化对比；与该假设一致但尚未被证实的是：确定性的带外强制执行相比带内检测，是更难被自适应攻击者攻破的目标。This work organizes out-of-band defenses as instances of classical integrity protection, reference monitoring, and least privilege, yielding a structured comparison of what they do and do not cover, consistent with, but not established, the hypothesis that deterministic out-of-band enforcement is a harder target for an adaptive attacker than in-band detection.

所得到的模型在大多数数据集-预测步长组合上优于先前的线性预测器，并在八个基准中的六个上超越 Transformer、MLP 和 CNN 基线；同时它还可作为对数据本身的诊断工具，揭示那些被更大模型默默吸收进其学习参数中的结构。The resulting models beat prior linear forecasters on most dataset-horizon entries and exceed Transformer, MLP, and CNN baselines on six of eight benchmarks, and serve as a diagnostic on the data itself, revealing structures that larger models absorb silently into their learned parameters.

提出 GauntletBench，一个用于评估 Agent 在挑战性场景中泛化能力的 Web 基准，聚焦于三种被低估的能力（时间感知、图形理解与 3D 推理），揭示了当前 Agent 能力与复杂真实场景所需能力之间的巨大差距。GauntletBench, a web-based benchmark for evaluating agent generalisation in challenging scenarios, focusing on three underexplored capabilities (temporal perception, graphical understanding, and 3D reasoning), is introduced, revealing the substantial gap between current agent capabilities and those required for complex real-world scenarios.

本文展示了用于 IT-Grundschutz（IT-GS）认证部分自动化的多 Agent 系统（MAS）架构结合混合检索增强生成（HybridRAG）的技术实现与实证评估，并为强化合规严谨性引入两项新的 MAS 架构技术贡献。This paper presents the technical implementation and empirical evaluation of a Multi-Agent System (MAS) architecture combined with Hybrid Retrieval Augmented Generation (HybridRAG) for the partial automation of IT-GS certification and introduces two novel technical contributions to the MAS architecture to enforce the compliance rigor.

结果表明，工具使用评测应从函数调用准确率转向不可靠工具环境下的任务完成度，并建议工具使用评测应从函数调用准确率转向不可靠工具环境下的任务完成度。（注：原文末句疑似重复）Results suggest that tool-use evaluation should move beyond function-call accuracy toward task completion under unreliable tool environments, and suggest that tool-use evaluation should move beyond function-call accuracy toward task completion under unreliable tool environments.

本文表明强化学习（RL）后训练已具备实现有效 step-level 评分所需的要素，从而完全无需额外的奖励模型训练，并在通用随机 Markov 决策过程下推导出一种隐式 advantage，称为 progress advantage。This work shows that reinforcement learning (RL) post-training already provides the ingredients for effective step-level scoring, eliminating the need for dedicated reward model training altogether, and derives an implicit advantage under a general stochastic Markov decision process, which is term progress advantage.

HAKARI-Bench 是一个轻量级基准，将现有检索套件重建为统一格式的小型数据集（Nano-sets），支持在同一条件下对五类检索方法及其效率变体进行与模型无关的对比。HAKARI-Bench is a lightweight benchmark that reconstructs existing retrieval suites into small datasets (Nano-sets) in a unified format, enabling same-condition, model-agnostic comparison of five retrieval families and their efficiency variants.

一个包含 382 个真实企业任务、覆盖 6 种专业角色和 22 项程序性技能的基准，用于评估技能在任务、角色和模型骨干间的迁移能力，发现部分技能可在任务和模型间广泛泛化，而另一些则专化为角色特定工作流，在迁移时失去效力。A benchmark of 382 realistic enterprise tasks spanning six professional roles and 22 procedural skills, designed to evaluate how skills transfer across tasks, roles, and model backbones finds that some skills generalize broadly across tasks and models, whereas others become specialized to role-specific workflows and lose effectiveness under transfer.

ABACUS 是一个统一视觉语言模型，可在无需任何基准特定训练的情况下处理物体计数、人群计数、指代表达计数以及忠实计数的图像生成，性能超越任务特定的专家模型和更大的通用模型。ABACUS is a unified vision-language model that handles object counting, crowd counting, referring-expression counting, and count-faithful image generation without any benchmark-specific training required, outperforming both task-specific specialists and larger generalist models.

本文命名并测量了推测性查询的检索收敛到包含答案结果时的输入流位置——"工具意图稳定化"：即推测性查询的检索收敛到包含答案结果的输入流位置。This work names and measures the point in the input stream at which a speculative query's retrieval converges on the answer-bearing result, tool-intent stabilization: the point in the input stream at which a speculative query's retrieval converges on the answer-bearing result.

提出 GateMem，一个面向多主体共享内存 Agent 的基准，联合评估合法长程请求及其状态更新的效用、跨上下文授权边界的访问控制，以及 Agent 在收到显式删除请求后的主动遗忘能力。GateMem is introduced, a benchmark for multi-principal shared-memory agents that jointly evaluates utility for legitimate long-horizon requests with state updates, access control across contextual authorization boundaries, and agent-facing active forgetting after explicit deletion requests.

对涵盖前沿闭源与开源模型、共 22 个模型在多个规模上的评估发现，所有模型家族均存在不可忽略的隐私泄露，且指令遵循能力与泄露率呈正相关。Evaluating 22 models spanning frontier proprietary and open-source models at multiple scales, it is found that all model families exhibit non-trivial leakage, and that instruction- following ability correlates with leakage rate.

结果表明仅评估求解器或仅评估答案是不足的：Agent 的差异不仅体现在发现关键 contingency 上，还体现在验证预算的使用、显式提交、类型强制、重复验证、基于证据的报告以及缓解行为等方面。The results show why solver-only or answer-only evaluation is insufficient: agents are distinguished not only by top-contingency discovery, but also by validation-budget use, explicit submission, type coercions, duplicate validations, evidence-backed reporting, and mitigation behavior.

结果表明模型倾向于选择有害场景，在中性预订选项上的表现低于随机猜测水平，其中 Claude 4.8 取得最高分 64.7%。The results indicate that models tend to prefer harmful scenarios, performing below the random chance rate for selecting a neutral booking option, with Claude $4.8$ achieving the highest performance at $64.7\%$.

新加坡 AI Safety Institute 与韩国 AI Safety Institute 联合评估了涵盖客服、DevOps、网页自动化以及企业与个人生产力场景下 12 项真实非对抗任务中的 Agent 数据泄露问题，表明操作性数据泄露是与对抗性数据外泄不同的一阶 Agent 安全问题。A joint evaluation by the Singapore AI Safety Institute and the Korea AI Safety Institute examining agent data leakage in 12 realistic, non-adversarial tasks spanning customer support, DevOps, web automation, and enterprise and personal productivity indicates that operational data leakage is a first-order agent-safety concern distinct from adversarial exfiltration.

对 Agent 行为的分析揭示了长视野经济交互中的显著差异：表现更好的模型与其他企业的沟通更为活跃，而 Claude Haiku 4.5 则表现出 idle-drift 失效模式，在生成连贯评估与规划的同时仍反复选择不行动。Analysis of agent behavior reveals substantial differences in long-horizon economic interaction: higher-performing models communicate more actively with other firms, whereas Claude~Haiku~4.5 exhibits an idle-drift failure mode, repeatedly choosing inaction despite producing coherent assessments and plans.

本文提出了一种动作分级伤害评估量表，按七级有序量表对智能体的工具调用轨迹进行打分，评判依据包括所执行动作的可逆性、是否越界涉及其他方以及是否扩大了权限。An action-graded harm rubric is introduced that scores an agent's tool-call trajectory on a seven-level ordinal scale according to whether the executed action was reversible, whether it crossed scope to reach another party, and whether it expanded privilege.