正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Is thumbnail A/B testing sequential or in parallel? Here’s the answer!

YouTube Creator Insider

39,216 subscribers

27,161 次观看 • 2 年前 •via X (Twitter)

科学技术教育

Anya Rossi• Live Now

Private livecam show

9 条评论

Anis 🇩🇿 的头像

Anis 🇩🇿2 年前

Will there be an option in the future to let the A/B testing pick the winning thumbnail in the first few hours?

YouTube Creator Insider 的头像

YouTube Creator Insider2 年前

The bounding factor isn’t time, it’s data. There has to be enough time to get enough data to pick the winner. If you were to pick an arbitrary time like the first few hours, there’s a likelihood you wouldn’t have enough data to pick a winner. Make sense?

Ed Andersen 的头像

Ed Andersen2 年前

What's the threshold for a thumbnail test to complete "early"? Ran one expecting it to take two weeks but a winner was decided in a day. Is it a certain number of impressions? Fixed percentage of watch time being hit?

YouTube Creator Insider 的头像

YouTube Creator Insider2 年前

Enough data to determine with statistical confidence

Nona Prince 的头像

Nona Prince2 年前

Pls add this for live streams also or atleast when live is done we can run the test

Michel Lomner 的头像

Michel Lomner2 年前

Why did you stop posting on YouTube?

Vinay Dhareshwar 的头像

Vinay Dhareshwar2 年前

Ah, great to know! Thanks for this, Rene. 🙌

Anis 🇩🇿 的头像

Anis 🇩🇿2 年前

If I use 3 thumbnails, will they target 3 different audiences, or is it just random?

Tim Schofield 的头像

Tim Schofield2 年前

Thanks for the answer!

相关视频

🚨 New YouTube Feature Alert 🚨 Title AND Thumbnail A/B testing is starting to roll out 🙌🏾

🚨 New YouTube Feature Alert 🚨 Title AND Thumbnail A/B testing is starting to roll out 🙌🏾

Marques Brownlee

334,710 次观看 • 1 年前

I asked 6 agents in two conditions (parallel v/s sequential) to create the same tower defence game 👇 LLMs are known to have mode collapse, i.e. if you have multiple parallel runs on the same prompt, you get very similar responses. Is there a way to increase diversity? Yes, it turns out that if you do such multiple runs **sequentially**, instead of in parallel. (Plus for each sequential run, append results from previous runs and instruct the LLM to create something different from before.) I did this quick experiment to explore parallel v/s sequential runs for diversity. Given the same prompt to build a tower defence game, sequential runs had a much higher diversity and innovation v/s parallel runs. Check out this (AI-generated) video for gameplay clips. The sequential cohort created interesting new dynamics and visual style, while parallel ones all collapsed to similar colors and orbital style dynamics.

I asked 6 agents in two conditions (parallel v/s sequential) to create the same tower defence game 👇 LLMs are known to have mode collapse, i.e. if you have multiple parallel runs on the same prompt, you get very similar responses. Is there a way to increase diversity? Yes, it turns out that if you do such multiple runs sequentially, instead of in parallel. (Plus for each sequential run, append results from previous runs and instruct the LLM to create something different from before.) I did this quick experiment to explore parallel v/s sequential runs for diversity. Given the same prompt to build a tower defence game, sequential runs had a much higher diversity and innovation v/s parallel runs. Check out this (AI-generated) video for gameplay clips. The sequential cohort created interesting new dynamics and visual style, while parallel ones all collapsed to similar colors and orbital style dynamics.

Paras Chopra

11,592 次观看 • 2 个月前

Is tolerance a utility in the region? Here’s the answer 👇

Is tolerance a utility in the region? Here’s the answer 👇

Rauda Altenaiji

21,699 次观看 • 2 个月前

Is there a future for Jews in the US? Stupid question. Here’s my answer:

Is there a future for Jews in the US? Stupid question. Here’s my answer:

Emily Austin

337,483 次观看 • 1 年前

Sequential Monte Carlo speculative decoding from Makora keeps multiple draft tokens alive in parallel instead of rewinding failed matches.

Sequential Monte Carlo speculative decoding from Makora keeps multiple draft tokens alive in parallel instead of rewinding failed matches.

SemiAnalysis

24,418 次观看 • 4 个月前

We're proud to present you the next generation sequential shifter: MagShift 🙌 Heusinkveld MagShift is a sequential shifter with a patented magnetic shift system. Feel the gears click into place like in real race cars. MagShift is the most versatile sequential shifter available.

We're proud to present you the next generation sequential shifter: MagShift 🙌 Heusinkveld MagShift is a sequential shifter with a patented magnetic shift system. Feel the gears click into place like in real race cars. MagShift is the most versatile sequential shifter available.

Heusinkveld

29,182 次观看 • 2 年前

Claude Sonnet 5 just leaked. If this is real, dev teams change overnight. Here’s why: → Massive context window (entire codebases). → Multi-agent workflows running in parallel. → Faster reasoning + fewer hallucinations. → Planning, coding, testing happening at once. → Infrastructure built for speed. This isn’t “ask → answer.” It’s delegation. Save this video, you’ll prepare properly. Want the SOP? DM me. 💬

Claude Sonnet 5 just leaked. If this is real, dev teams change overnight. Here’s why: → Massive context window (entire codebases). → Multi-agent workflows running in parallel. → Faster reasoning + fewer hallucinations. → Planning, coding, testing happening at once. → Infrastructure built for speed. This isn’t “ask → answer.” It’s delegation. Save this video, you’ll prepare properly. Want the SOP? DM me. 💬

Julian Goldie SEO

19,731 次观看 • 8 个月前

Parallel Unique Avatars Marcolian // Mirielle In game testing

Parallel Unique Avatars Marcolian // Mirielle In game testing

Parallel Sanctuary

18,435 次观看 • 1 年前

I was asked if Reform is left or right wing. Here’s my answer:

I was asked if Reform is left or right wing. Here’s my answer:

Zia Yusuf

109,486 次观看 • 1 年前

🚨Silver near $70 an ounce!! Is it just a “fad” or a “craze”? Here’s my answer! Please share this to spread the word! ‼️

🚨Silver near $70 an ounce!! Is it just a “fad” or a “craze”? Here’s my answer! Please share this to spread the word! ‼️

Bald Guy Money

45,559 次观看 • 10 个月前

Today we made one of our internal tools at Big Bang public and it’s free to use for #RobloxDev ✴️Big Bang Vision is a thumbnail testing tool that helps you see if your game thumbnail actually stands out. 📌Available in

Today we made one of our internal tools at Big Bang public and it’s free to use for #RobloxDev ✴️Big Bang Vision is a thumbnail testing tool that helps you see if your game thumbnail actually stands out. 📌Available in

Star ⭐️

13,916 次观看 • 5 个月前

Ish, Flip, and Ice, answer where they rank (A, B, or C) in the urban podcast world. Where would you rank the 3? #JBTVClips

Ish, Flip, and Ice, answer where they rank (A, B, or C) in the urban podcast world. Where would you rank the 3? #JBTVClips

JBTV Community

10,943 次观看 • 4 个月前

Ludwig reached out to MrBeast for his thumbnail data to help streamers improve their YouTube videos 👏❤️ “This comes directly from MrBeast, I asked him for his A/B testing.”

Ludwig reached out to MrBeast for his thumbnail data to help streamers improve their YouTube videos 👏❤️ “This comes directly from MrBeast, I asked him for his A/B testing.”

bloom 𖧧

214,165 次观看 • 3 个月前

In brief, architecture determines outcomes. When networks rely on sequential blocks, speed and efficiency hit limits. When consensus is designed for parallel processing, those limits disappear. 👉 hashgraph vs blockchain Learn more:

In brief, architecture determines outcomes. When networks rely on sequential blocks, speed and efficiency hit limits. When consensus is designed for parallel processing, those limits disappear. 👉 hashgraph vs blockchain Learn more:

Hedera

38,475 次观看 • 9 个月前

What's the deal with A/B testing? 🤔 Find out in the next Judo Talks, where Judo and David answer your questions about Clash events! Live July 28, 12 PM UTC on the Judo Sloth 2 channel.

What's the deal with A/B testing? 🤔 Find out in the next Judo Talks, where Judo and David answer your questions about Clash events! Live July 28, 12 PM UTC on the Judo Sloth 2 channel.

Clash of Clans

94,956 次观看 • 2 个月前

Listen to Samar Khanna explain why parallel generation, rather than sequential, raises the performance ceiling for language models. Learn more about diffusion LLMs. → We're hiring:

Listen to Samar Khanna explain why parallel generation, rather than sequential, raises the performance ceiling for language models. Learn more about diffusion LLMs. → We're hiring:

Inception

18,582 次观看 • 8 个月前