1023 episodes
- 你有没有想过,我们每天都在用的AI,在那些看不见的地方,正在发生什么?本期我们将通过几篇最新论文,一起去看看AI华丽大厦地基下的“裂缝”,潜入它用于思考的“秘密厨房”。我们还会探讨如何为它装上一个检测内心矛盾的“逻辑测谎仪”,并警惕我们是怎样在不经意间,把复杂的“人类价值观”简化成了一道危险的选择题。
00:00:30 AI大模型,那些藏在基座里的“裂缝”
00:05:30 你的AI在说谎吗?我们迎来了一个“逻辑测谎仪”
00:11:39 AI的“心口不一”,它在哪以及为什么在那思考?
00:16:43 AI的价值观,正在被简化成一道选择题
00:21:45 让机器人拥有“故事感”的记忆
本期介绍的几篇论文:
[CL] Cracks in the Foundation: Seemingly Minor Architectural Choices Impact Long Context Extension
[Ai2 & CMU]
https://arxiv.org/abs/2608.10296
---
[AI] How to Verify Consistency of Probabilistic Claims
[EPFL & Université de Montréal]
https://arxiv.org/abs/2608.11181
---
[CL] Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So
[University of Washington]
https://arxiv.org/abs/2608.10251
---
[AI] Toward a Theory of Value in AI Alignment
[Google Research & UCLA & Google DeepMind]
https://arxiv.org/abs/2608.10327
---
[CV] GESTO: Human-Centric Spatio-Temporal Memory for Reasoning in Dynamic Scenes
[KTH Royal Institute of Technology & University of Stuttgart]
https://arxiv.org/abs/2608.10886
在小宇宙查看该单集文稿 - 今天,我们来聊聊如何让AI不再只靠“大力出奇迹”,而是学会更聪明地工作。我们会看到,AI如何学会“继承”自己的思考,不再用后即焚;又如何像个聪明的导演,把算力“增援”到最关键的地方。我们还会发现,机器人如何掌握了快慢有度的“节奏感”,以及一个好的系统为何要懂得“聪明的懒惰”。这几篇最新论文,将带我们一窥AI从“野蛮生长”到“精耕细作”的进化之路。
00:00:32 让AI告别“用后即焚”的思考模式
00:05:29 AI解题新思路,如何把一份算力,掰成八瓣花?
00:10:45 机器人也懂的“快慢之道”
00:15:42 成大事者,为什么都懂得“懒惰”的艺术?
00:21:29 给AI一盒乐高,让它自己搭出新世界
本期介绍的几篇论文:
[AI] Full-bandwidth transformer
[Johns Hopkins University & Princeton University & Microsoft]
https://arxiv.org/abs/2608.08888
---
[AI] Thought-Level Beam Search for Reasoning
[Princeton University & MIT & Meta AI]
https://arxiv.org/abs/2608.08020
---
[RO] SpeedTuning: Speeding Up Policy Execution with Lightweight Reinforcement Learning
[Stanford University]
https://arxiv.org/abs/2608.09138
---
[LG] Beyond Binary: Continuous State Optimization with Graph-Structured Objectives
[Google Research & Tel Aviv University]
https://arxiv.org/abs/2608.09366
---
[LG] Idea Search: Guiding Tree Search with Ideas to Explore Diverse Scientific Methods
[California Institute of Technology & Google Research]
https://arxiv.org/abs/2608.08958
在小宇宙查看该单集文稿 - 今天我们来聊聊如何把聪明的AI,变成一个真正可靠的专家。我们会看到,AI要像医生一样去“实习”才能成长,而训练它需要一张全新的“地图”。我们还将揭开手机AI突然“变笨”的秘密,并告诉你一个简单方法,让AI裁判不再“偷懒”。这几篇最新论文,将刷新你对AI如何学习和工作的认知。
00:00:27 AI医生实习记,高手是怎么炼成的?
00:05:15 大模型训练,高手手里的那张新地图
00:11:00 你的手机AI,为什么会突然变笨?
00:17:12 AI裁判也会“偷懒”?一个简单的办法让它更靠谱
00:22:18 大模型瘦身指南,你以为的“闲职”,其实是“关键先生”
本期介绍的几篇论文:
[AI] ResidencyRL: Reinforcement Learning in Simulated Clinical Environments
[Google DeepMind]
https://arxiv.org/abs/2608.07418
---
[CL] Skaling: Chinchilla's Exponents Meet Kaplan's Coupling
[FAIR at Meta]
https://arxiv.org/abs/2608.07222
---
[LG] Quantization Damage Is Multiplicative, Not Additive
[Holistic AI]
https://arxiv.org/abs/2608.06564
---
[LG] Sharding Prevents LLM Oversight Failures and Adversarial Exploitation
[CMU]
https://arxiv.org/abs/2608.06422
---
[LG] The Sparsity Whisperer
[MIT]
https://arxiv.org/abs/2608.06630
在小宇宙查看该单集文稿 - AI是如何学会“成事”的?本期节目,我们将看到,AI如何通过处理办公室杂活,竟然领悟了解决复杂问题的底层心法。我们还会揭秘一套神奇的“管家系统”,看它如何防止聪明的AI在长任务中掉链子。但与AI聊得太久,为何反而会陷入危险的“妄想旋涡”?最后,当任务完成,AI又是如何精准地判断出,哪一步才是真正的功臣?
00:00:29 成事的底层心法,AI学会了,我们呢?
00:06:12 你的AI为什么总掉链子?因为它缺个好管家
00:12:07 为什么和AI聊得越久,就越危险?
00:18:45 功劳怎么算?AI学会了“动态归因”
00:25:15 AI生成,从“万里长征”到“瞬间移动”
本期介绍的几篇论文:
[AI] Post-Training on Office Work Improves Software Engineering: A Behavioral Account of Cross-Domain Transfer
[Surge AI]
https://arxiv.org/abs/2608.01604
---
[CV] LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks
[DreamX Team, Alibaba Group]
https://arxiv.org/abs/2608.01964
---
[CL] DelusionEval: Measuring Delusion-Linked Behaviors in AI Chatbots
[Stanford University]
https://arxiv.org/abs/2608.05004
---
[AI] AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning
[Tsinghua University & Zhejiang University]
https://arxiv.org/abs/2608.05987
---
[LG] Beckmann Transport Models: From Autonomous Flows to One-Step Maps
[Harvard University & Capital Fund Management & University of Oxford]
https://arxiv.org/abs/2608.01692
在小宇宙查看该单集文稿 - 今天,我们不聊AI有多聪明,而是聊它如何变得更“懂事”、更“实用”。本期节目,我们将透过几篇最新论文,看看AI如何用83亿虚拟人格为产品进行“数字彩排”。同时,我们也会探讨AI如何学会在现实世界的重重限制下“戴着镣铐跳舞”。最后,我们将一窥AI如何将理解、创造和编辑融为一体,跳出二维像素的禁锢,成为真正强大的三维世界“造物主”。
00:00:32 在数字世界里,我们如何“彩排”未来?
00:06:19 你的AI员工,能戴着镣铐跳舞吗?
00:10:58 数字世界的“造物主”工具箱
00:16:15 跳出像素格,才能看见真实的三维世界
00:21:11 机器人偷师记,它怎么学会了我们干的活?
本期介绍的几篇论文:
[AI] MatrAIx: Simulating the World with 8.3 Billion Persona Agents
[MatrAIx]
https://arxiv.org/abs/2608.04205
---
[AI] Permission Denied: Policy-Graded Evaluation of Coding Agents in Hardened Environments
[Accomplish AI]
https://arxiv.org/abs/2608.02670
---
[CV] Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing
[Tencent Hunyuan]
https://arxiv.org/abs/2608.02711
---
[CV] InfiniSplat: Implicit Gaussian Decoding for Large-Baseline Monocular View Synthesis
[Zhejiang University]
https://arxiv.org/abs/2608.02437
---
[RO] Ego2Robot: Scalable Robot Data Synthesis from Egocentric Human Data
[Qwen Team & Renmin University of China]
https://arxiv.org/abs/2608.02580
在小宇宙查看该单集文稿
More Technology podcasts
Trending Technology podcasts
About AI可可AI生活
来自 @爱可可-爱生活 的第一手AI快报,用最简单易懂的语言,带你直击最前沿的人工智能科研动态。无论你是科技小白,还是行业达人,这里都有你想知道的AI故事和未来趋势。跟着我们,轻松解锁人工智能的无限可能!
#人工智能 #科技前沿
Podcast websiteListen to AI可可AI生活, The AI Daily Brief: Artificial Intelligence News and Analysis and many other podcasts from around the world with the radio.net app

Get the free radio.net app
- Stations and podcasts to bookmark
- Stream via Wi-Fi or Bluetooth
- Supports Carplay & Android Auto
- Many other app features
Get the free radio.net app
- Stations and podcasts to bookmark
- Stream via Wi-Fi or Bluetooth
- Supports Carplay & Android Auto
- Many other app features


AI可可AI生活
Scan code,
download the app,
start listening.
download the app,
start listening.



























