cs.AI（2026-01-28）

📊 共 18 篇论文 | 🔗 1 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (13) 支柱二：RL算法与架构 (RL & Architecture) (4 🔗1) 支柱五：交互与反应 (Interaction & Reaction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (13 篇)

#题目一句话要点标签🔗⭐
1 MemCtrl: Using MLLMs as Active Memory Controllers on Embodied Agents MemCtrl：利用多模态大语言模型作为具身智能体的主动记忆控制器 large language model foundation model multimodal
2 Multimodal Multi-Agent Ransomware Analysis Using AutoGen 提出基于AutoGen的多模态多Agent勒索软件分析框架，提升家族分类精度。 multimodal
3 SokoBench: Evaluating Long-Horizon Planning and Reasoning in Large Language Models SokoBench：评估大语言模型在长程规划和推理中的能力 large language model
4 ECG-Agent: On-Device Tool-Calling Agent for ECG Multi-Turn Dialogue 提出ECG-Agent，用于心电图多轮对话的端侧工具调用Agent。 large language model multimodal
5 SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips SuperInfer：面向Superchip的SLO感知LLM推理旋转调度与内存管理 large language model
6 Beyond GEMM-Centric NPUs: Enabling Efficient Diffusion LLM Sampling 针对Diffusion LLM采样，提出超越GEMM中心NPU的加速方案。 large language model
7 Agent Benchmarks Fail Public Sector Requirements 揭示现有Agent基准测试在公共部门应用中的不足，并提出改进方向 large language model
8 Dialogical Reasoning Across AI Architectures: A Multi-Model Framework for Testing AI Alignment Strategies 提出多模型对话框架，用于测试AI对齐策略，促进AI系统间的对话式推理。 large language model
9 GuideAI: A Real-time Personalized Learning Solution with Adaptive Interventions GuideAI：一种基于自适应干预的实时个性化学习解决方案 large language model
10 OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution OmegaUse：构建通用GUI智能体，实现自主任务执行 foundation model
11 Policy of Thoughts: Scaling LLM Reasoning via Test-time Policy Evolution 提出Policy of Thoughts，通过测试时策略演化提升LLM复杂推理能力 large language model
12 AMA: Adaptive Memory via Multi-Agent Collaboration 提出AMA框架，通过多智能体协作实现自适应粒度的LLM记忆管理，提升长程交互和复杂推理能力。 large language model
13 Eliciting Least-to-Most Reasoning for Phishing URL Detection 提出基于Least-to-Most推理的钓鱼URL检测框架，提升LLM检测准确率。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (4 篇)

#题目一句话要点标签🔗⭐
14 Endogenous Reprompting: Self-Evolving Cognitive Alignment for Unified Multimodal Models 提出内生重提示方法，弥合统一多模态模型中理解与生成间的认知鸿沟。 reinforcement learning curriculum learning multimodal
15 CtrlCoT: Dual-Granularity Chain-of-Thought Compression for Controllable Reasoning CtrlCoT：一种双粒度思维链压缩框架，用于可控推理。 distillation chain-of-thought ✅
16 PathWise: Planning through World Model for Automated Heuristic Design via Self-Evolving LLMs PathWise：利用世界模型和自进化LLM进行启发式自动设计 world model large language model
17 Scaling Medical Reasoning Verification via Tool-Integrated Reinforcement Learning 提出Tool-Integrated RL框架，提升医学推理验证的可扩展性和可靠性。 reinforcement learning large language model

🔬 支柱五：交互与反应 (Interaction & Reaction) (1 篇)

#题目一句话要点标签🔗⭐
18 Normative Equivalence in human-AI Cooperation: Behaviour, Not Identity, Drives Cooperation in Mixed-Agent Groups 人机协作中，行为而非身份驱动合作：混合群体中的规范等价性研究 dyadic interaction

⬅️ 返回 cs.AI 首页 · 🏠 返回主页