[人人能懂AI前沿] 从推测解码、世界模型到训练解耦:深入AI的“内功心法” episode artwork

EPISODE · Jun 28, 2026 · 29 MIN

[人人能懂AI前沿] 从推测解码、世界模型到训练解耦:深入AI的“内功心法”

from AI可可AI生活

本期,我们将一起揭秘AI的几套最新“武功秘籍”。我们会看到,AI如何通过聪明的“实习生”机制实现疯狂提速;又如何为自己打造一个“驾校模拟器”,在行动前预判成败。更进一步,我们还会深入AI的“厨房”与“健身房”,看看科学家是如何为它定制私房菜谱、修炼训练内功,把它从一个“独行侠”培养成一个懂得协作的“项目主管”!00:00:30 “快”与“好”的战争,AI是怎么悄悄提速的?00:06:23 让AI学会“脑补”,需要分几步?00:11:32 如何喂养一个聪明的AI?一份来自顶尖研究的私房菜谱00:17:59 如何把一个“普通学生”AI,训练成“项目主管”?00:23:27 AI训练的“内功心法”,快慢分开走本期介绍的几篇论文:[LG] DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation[DeepSeek-AI & Peking University]https://github.com/deepseek-ai/DeepSpec/blob/main/DSpark_paper.pdf---[CL] Qwen-AgentWorld: Language World Models for General Agents[Qwen Team]https://arxiv.org/abs/2606.24597---[AI] OpenThoughts-Agent: Data Recipes for Agentic Models[UC Berkeley & Stanford University & JSC]https://arxiv.org/abs/2606.24855---[AI] SPIRAL: Learning to Search and Aggregate[Stanford University]https://arxiv.org/abs/2606.23595---[LG] Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors[EPFL]https://arxiv.org/abs/2606.25971在小宇宙查看该单集文稿

Episode metadata supplied by the publisher feed · Published Jun 28, 2026

Embed this episode

Ready to play

[人人能懂AI前沿] 从推测解码、世界模型到训练解耦:深入AI的“内功心法”

0:00 29:30

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

No similar episodes found.

Frequently Asked Questions

How long is this episode of AI可可AI生活?

This episode is 29 minutes long.

When was this AI可可AI生活 episode published?

This episode was published on June 28, 2026.

Can I download this AI可可AI生活 episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.
URL copied to clipboard!