Class Activation Mapping in Explainable Computer Vision: A Method-Centered Review of CNN, Transformer, and Foundation-Model-Era Visual Explanations

系统综述类激活映射（CAM）在可解释计算机视觉中的方法演进，以方法为中心梳理CNN、Transformer及基础模型时代的视觉解释技术。比较各类CAM变体的原理、适用场景与局限性，揭示解释方法的代际变迁，为该领域研究提供系统性参考和未来发展方向指引。

📅 08-12📎 arXiv AI Papers

AI Coding Workflows: From Cloud to Local

AI 周报，深度分析行业趋势：AI Coding Workflows: From Cloud to Local…

📎 The Batch

DreamFly: Causal Memory and Receding-Horizon Diffusion Planning for Aerial Vision-Language Navigation

针对空中视觉语言导航任务，提出DreamFly框架，融合因果记忆与滚动时域扩散规划。该框架有效捕捉历史决策因果关联，并结合动态扩散模型进行长程轨迹规划，显著提升无人机在复杂三维环境中根据自然语言指令导航的准确性与鲁棒性。

📅 08-12📎 arXiv AI Papers

AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses

提出一种测试时强到弱能力迁移方法AI4AI，通过”harness”机制将强模型能力在推理阶段传递给弱模型，无需重新训练。该方法利用强模型输出作为辅助信号引导弱模型推理，有效提升弱模型的复杂任务表现，为模型能力增强提供新思路。

📅 08-12📎 arXiv AI Papers

Redistribution-based Cost Inference Improves Sparse Safe Offline RL

提出基于重分配的代价推断方法，改进稀疏安全离线强化学习。该方法从稀疏回报中重新分配代价信号，构建更稠密的安全约束估计，缓解稀疏奖励环境下代价估计不准确的问题，提升策略学习的样本效率与安全性，为安全关键场景提供可行方案。

📅 08-12📎 arXiv AI Papers

Constructing Dynamic Master Logic Models as Knowledge Graphs for Complex System Diagnostics Using Retrieval-Augmented Large Language Models

提出利用检索增强大语言模型构建动态主逻辑模型知识图谱的新方法，用于复杂系统诊断。该方法可自动整合系统文档与运行数据，动态更新逻辑关系与因果链路，提升故障根因定位的准确性和可解释性，为复杂装备智能运维提供高效技术支撑。

📅 08-12📎 arXiv AI Papers

Beyond Trial-and-Error: Agentic Optimization for Image-to-Video Adherence

提出超越试错范式的智能体优化框架，用于图像到视频生成的一致性问题。通过智能体自主规划、执行与反思迭代，动态优化生成策略，提升输出视频对输入图像的语义和视觉保真度，显著降低人工调参成本，实现跨模态对齐的高效精准优化。

📅 08-12📎 arXiv AI Papers

VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies

Agents deployed in enterprise settings must reason across structured APIs and document collections, yet existing benchmarks evaluate these capabilities in isolation. We introduce VAKRA (e\textbf{V}aluating \textbf{A}PI and \textbf{K}nowledge \textbf{R}etrieval \textbf{A}gents), a benchmark of over $

📅 08-12📎 arXiv AI Papers

Structural Silence: When AI Infrastructure Fails Speakers of Underrepresented Languages

Artificial intelligence tools for education and language support are increasingly framed as scalable responses to access gaps in under-resourced communities. Yet the infrastructure underlying these tools, including training corpora, tokenization schemes, evaluation benchmarks, and deployment archite

📅 08-12📎 arXiv AI Papers

A Neighborhood Attention Transformer Network for Enhanced 3D Segmentation of the Left Anterior Descending Artery

Background: Accurate segmentation of the Left Anterior Descending (LAD) artery in 3D free-breathing, non-contrast CT is critical for cardiac dose sparing in thoracic radiotherapy. The LAD is extremely small, has poor soft-tissue contrast, and varies substantially across patients; even manual contour

📅 08-12📎 arXiv AI Papers

Convergent Detour Hijacking: Task-Preserving Resource Amplification in Skill-Based LLM Agents

提出一种针对基于技能的LLM智能体的新攻击范式“汇合绕行劫持”，通过保留任务目标的同时放大资源消耗，使智能体陷入冗余执行路径。该方法利用技能组合漏洞，在不破坏任务完成度的情况下显著提升计算开销，揭示了当前智能体安全机制在资源控制层面的薄弱环节，为构建健壮的代理系统提供了重要防御视角。

📅 08-12📎 arXiv AI Papers

Empty shelves or lost keys? Recall is the bottleneck for parametric factuality

本文探讨大语言模型参数化事实性的失败根源，指出主要瓶颈在于“回忆”能力而非知识存储缺失。正如超市货架并非为空，而是钥匙丢失难以取出物品，模型虽已掌握信息却无法在需要时准确调用。研究通过实验区分存储与访问差异，并提出针对性改进方向。

📅 08-12📎 Google AI Blog