2026.01.15 | 算法自进化夺冠;LLM远瞻省token

2026.01.15 | 算法自进化夺冠;LLM远瞻省token

11分钟 ·
播放数147
·
评论数0

本期的 15 篇论文如下:

00:20 🧬 Controlled Self-Evolution for Algorithmic Code Optimization(用于算法代码优化的受控自进化方法)

00:52 🧠 MAXS: Meta-Adaptive Exploration with LLM Agents(MAXS:基于大语言模型智能体的元自适应探索)

01:27 🧠 $A^3$-Bench: Benchmarking Memory-Driven Scientific Reasoning via Anchor and Attractor Activation(A³-Bench:通过锚点与吸引子激活基准测试记忆驱动的科学推理)

02:10 🔍 DeepResearchEval: An Automated Framework for Deep Research Task Construction and Agentic Evaluation(DeepResearchEval:面向深度研究任务构建与智能体评估的自动化框架)

02:53 🔬 SkinFlow: Efficient Information Transmission for Open Dermatological Diagnosis via Dynamic Visual Encoding and Staged RL(SkinFlow:通过动态视觉编码与分阶段强化学习实现开放皮肤病诊断的高效信息传输)

03:49 ⚡ Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning(Fast-ThinkAct:基于可言语化潜在规划的高效视觉-语言-动作推理)

04:20 🧊 OpenVoxel: Training-Free Grouping and Captioning Voxels for Open-Vocabulary 3D Scene Understanding(OpenVoxel:无需训练的体素分组与描述,实现开放词汇3D场景理解)

05:03 🧠 Distribution-Aligned Sequence Distillation for Superior Long-CoT Reasoning(面向卓越长链思维推理的分布对齐序列蒸馏)

06:04 🧠 ExpSeek: Self-Triggered Experience Seeking for Web Agents(ExpSeek:面向网络智能体的自触发经验寻求方法)

06:53 ⚠ Are LLMs Vulnerable to Preference-Undermining Attacks (PUA)? A Factorial Analysis Methodology for Diagnosing the Trade-off between Preference Alignment and Real-World Validity(大型语言模型是否易受偏好颠覆攻击?一种诊断偏好对齐与现实有效性权衡的因子分析方法论)

07:30 🔄 EvoFSM: Controllable Self-Evolution for Deep Research with Finite State Machines(EvoFSM:基于有限状态机的可控自演化深度研究框架)

08:04 🧠 Imagine-then-Plan: Agent Learning from Adaptive Lookahead with World Models(想象而后规划:基于世界模型的自适应前瞻智能体学习)

08:46 🌐 TranslateGemma Technical Report(TranslateGemma技术报告)

09:22 🧠 The AI Hippocampus: How Far are We From Human Memory?(AI海马体:我们距离人类记忆还有多远?)

10:03 🎯 FocusUI: Efficient UI Grounding via Position-Preserving Visual Token Selection(FocusUI:通过位置保持的视觉令牌选择实现高效用户界面定位)

【关注我们】

您还可以在以下平台找到我们,获得播客内容以外更多信息

小红书: AI速递