EPISODE · Jun 16, 2026 · 7 MIN
AI计算广告论文播报 2026-06-15：出价的'平均陷阱'与改写的稳定性闸门
from 周六9点半
本期聚焦商业化系统从'能用'到'经得起线上摔打'的工程化升级，三个关键词：奖励对齐、稳定性闸门、加性兜底。本期重点* 分布式检索增强出价（DRIVE: Distributional and Retrieval-Augmented Bidding with Value Evaluation）：揭露自动出价里隐蔽的'平均动作陷阱'——同状态下高低两种有效策略被回归头平均成谁也不像的中间价。论文用GMM多峰头+历史高回报检索+IQL critic的三件套即插即用解决，单步推断仅11ms，能挂在DT/CDT/PDiT等主流骨干上。* 持续奖励微调的查询改写器（CoRe: A Continuously Reward-Finetuned LLM Query Rewriter）：TikTok短视频搜索周更5个月的工业实录。最值得抄的两点是奖励函数的乘法结构与线上融合公式严格对齐，以及上线闸门同时盯...去小宇宙查看完整单集简介在小宇宙查看该单集文稿
Embed this episode
Ready to play
AI计算广告论文播报 2026-06-15：出价的'平均陷阱'与改写的稳定性闸门
0:00
7:21
1×
No transcript for this episode yet
Similar Episodes
No similar episodes found.
Similar Podcasts
No similar podcasts found.
Frequently Asked Questions
How long is this episode of 周六9点半?
This episode is 7 minutes long.
When was this 周六9点半 episode published?
This episode was published on June 16, 2026.
Can I download this 周六9点半 episode?
Yes. Use the download control on the episode player to save the publisher-provided media file.
URL copied to clipboard!