Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Complex instruction following is critical for LLM agents and applications. IOPO with notable improvements is proposed to consider both input and output preference pairs , not only aligning with response preferences but also meticulously exploring the instruction preferences.

Tongyi Lab

25,478 subscribers

1,762,955 views • 1 year ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

New short course on Reinforcement Learning from Human Feedback! RLHF is one of the key techniques that led to the rise of modern LLMs. It is used to align LLMs with human preferences, to make them more honest, helpful and harmless, by (i) learning a reward function that mimics human preferences, as expressed in human-provided labels, then, (ii) tuning an LLM to generate outputs that receive a high reward. In this course, taught by Nikita Namjoshi, Developer Advocate for GenAI at Google Cloud, you'll learn the details of how RLHF works, including how to apply it to tune an LLM for your own applications. You'll also use an open source library to tune a base LLM to align with human preferences expressed in a training set, and evaluate the tuned model by comparing its responses before and after RLHF-tuning. Please sign up here!

New short course on Reinforcement Learning from Human Feedback! RLHF is one of the key techniques that led to the rise of modern LLMs. It is used to align LLMs with human preferences, to make them more honest, helpful and harmless, by (i) learning a reward function that mimics human preferences, as expressed in human-provided labels, then, (ii) tuning an LLM to generate outputs that receive a high reward. In this course, taught by Nikita Namjoshi, Developer Advocate for GenAI at Google Cloud, you'll learn the details of how RLHF works, including how to apply it to tune an LLM for your own applications. You'll also use an open source library to tune a base LLM to align with human preferences expressed in a training set, and evaluate the tuned model by comparing its responses before and after RLHF-tuning. Please sign up here!

Andrew Ng

205,542 views • 2 years ago

Natural language instruction PLUS deterministic execution is the real unlock for AI agents that work consistently. With Headless 360, you can build these agents with the tools you already use: Claude Code, Cursor, Codex, Windsurf, Agentforce Vibes, Agentforce builder, and more.

Natural language instruction PLUS deterministic execution is the real unlock for AI agents that work consistently. With Headless 360, you can build these agents with the tools you already use: Claude Code, Cursor, Codex, Windsurf, Agentforce Vibes, Agentforce builder, and more.

Salesforce

1,783,451 views • 3 months ago

Don’t let the spirit of offense get the wrong message. Listen to people with your heart and not your hurt. I know we all want to defend and protect ourselves, but everyone is not out to hurt you. Let’s heal and take the correction, and instruction, and grow.

Don’t let the spirit of offense get the wrong message. Listen to people with your heart and not your hurt. I know we all want to defend and protect ourselves, but everyone is not out to hurt you. Let’s heal and take the correction, and instruction, and grow.

Erica Campbell

21,337 views • 2 years ago

Yes, America’s power grid is awesome 💪🏼 But speed to power and customer preferences matter too What if new customers would rather not deal with monopolies? Here’s the policy paper EEI doesn’t want you to read Freedom is the American way 🇺🇸

Yes, America’s power grid is awesome 💪🏼 But speed to power and customer preferences matter too What if new customers would rather not deal with monopolies? Here’s the policy paper EEI doesn’t want you to read Freedom is the American way 🇺🇸

Travis Fisher

21,791 views • 1 month ago

This is the last chatbot you’ll ever need. Yesterday, Mckay Wrigley built an oss Not Diamond-powered chat app. We loved it. So today we’re releasing a hosted version. Get the best LLM on every message and hyper-personalize routing to your preferences with feedback. Watch how:

This is the last chatbot you’ll ever need. Yesterday, Mckay Wrigley built an oss Not Diamond-powered chat app. We loved it. So today we’re releasing a hosted version. Get the best LLM on every message and hyper-personalize routing to your preferences with feedback. Watch how:

Tomas Hernando Kofman

19,668 views • 2 years ago

$Can someone actually parse the word salad of “Women are only allowed to want what gives men the ability to control them”. What does that even mean? I believe it is actually the reverse, women’s preferences and standards are accepted or receive a fraction of the pushback men’s preferences and standards receive. Men’s preferences are frequently pathologized and demonized to nuclear levels. Want a woman with little or no body hair? Want a young ADULT woman? You’re now instantly a p3d0.$

Can someone actually parse the word salad of “Women are only allowed to want what gives men the ability to control them”. What does that even mean? I believe it is actually the reverse, women’s preferences and standards are accepted or receive a fraction of the pushback men’s preferences and standards receive. Men’s preferences are frequently pathologized and demonized to nuclear levels. Want a woman with little or no body hair? Want a young ADULT woman? You’re now instantly a p3d0.

Brian Atlas

45,546 views • 11 months ago

🔥 We fine-tuned Gemma 3 using Unsloth AI on a custom instruction dataset scraped and built with . The model is now able to answer questions about OpenAI Agents SDK documentation. Check out the demo below👇

🔥 We fine-tuned Gemma 3 using Unsloth AI on a custom instruction dataset scraped and built with . The model is now able to answer questions about OpenAI Agents SDK documentation. Check out the demo below👇

Eric Ciarla (hiring)

28,794 views • 1 year ago

YC W24's nuvi.dev is a testing and simulation stack to harden LLM-based applications. AI teams leverage Relari to stress test mission-critical GenAI use cases with modular evaluation and synthetic data. Congrats on the launch, Yi Zhang and Pasquale!

YC W24's nuvi.dev is a testing and simulation stack to harden LLM-based applications. AI teams leverage Relari to stress test mission-critical GenAI use cases with modular evaluation and synthetic data. Congrats on the launch, Yi Zhang and Pasquale!

Y Combinator

14,312 views • 2 years ago

This is Neuro Sama, the first of it's kind AI VTuber and streamer, which not only is playing the game by itself but also interacting with chat... It requires moderation to keep in check but is constantly learning and getting better at both... holy

This is Neuro Sama, the first of it's kind AI VTuber and streamer, which not only is playing the game by itself but also interacting with chat... It requires moderation to keep in check but is constantly learning and getting better at both... holy

Jake Lucky

3,765,689 views • 3 years ago

Our first Generative AI short course in JavaScript! GitHub recently reported that JavaScript is again the world’s most popular programming language. To support web developers exploring and developing with generative AI, we just launched a new short course in JavaScript taught by Jacob Lee, founding engineer at . In Build LLM Apps with LangChain.js you’ll learn elements common in AI development, including: (i) Using data loaders to pull data from common sources such as PDFs, websites, and databases (ii) Prompts, which are used to provide the LLM context (iii) Modules to support RAG such as text splitters and integrations with vector stores (iv) Working with different models to write applications that are not vendor-specific (v) Parsers, which extract and format the output for your downstream code to process You’ll also build with the LangChain Expression Language, which lets you easily compose sequences (also called chains) of modules to perform complex tasks using LLMs. Putting all this together, you’ll also work on a conversational question-answering LLM application capable of using external data as context. Please sign up here:

Our first Generative AI short course in JavaScript! GitHub recently reported that JavaScript is again the world’s most popular programming language. To support web developers exploring and developing with generative AI, we just launched a new short course in JavaScript taught by Jacob Lee, founding engineer at . In Build LLM Apps with LangChain.js you’ll learn elements common in AI development, including: (i) Using data loaders to pull data from common sources such as PDFs, websites, and databases (ii) Prompts, which are used to provide the LLM context (iii) Modules to support RAG such as text splitters and integrations with vector stores (iv) Working with different models to write applications that are not vendor-specific (v) Parsers, which extract and format the output for your downstream code to process You’ll also build with the LangChain Expression Language, which lets you easily compose sequences (also called chains) of modules to perform complex tasks using LLMs. Putting all this together, you’ll also work on a conversational question-answering LLM application capable of using external data as context. Please sign up here:

Andrew Ng

284,320 views • 2 years ago

Today is a big day! We're launching Lovable Aesthetics 🔥 We just supercharged Lovable's design output. You can now ask for typography, layout, and color preferences, get various design concepts while building, and create much bolder landing pages, apps, and blogs. It's been my favorite thing to play with for the last few weeks. Go check it out!

Today is a big day! We're launching Lovable Aesthetics 🔥 We just supercharged Lovable's design output. You can now ask for typography, layout, and color preferences, get various design concepts while building, and create much bolder landing pages, apps, and blogs. It's been my favorite thing to play with for the last few weeks. Go check it out!

Felix Haas

104,846 views • 2 months ago

This clumsy response will haunt the new Minister of Higher Education, Buti Manamela, and his DG. The Skills Development Act is clear: “(4) If a SETA has not complied with an instruction issued in terms of section 14A, the Minister may- (b) invoke section 15 without further notice to the SETA.” In other words, the power to invoke section 15 flows from section 14A. Reliance on section 15(4) applies only to bypass consultation, in 15(1), not to irrationally skip issuing instructions. The Act requires the Minister to issue an instruction before invoking section 15. This is basic Administrative Justice. Basic!

This clumsy response will haunt the new Minister of Higher Education, Buti Manamela, and his DG. The Skills Development Act is clear: “(4) If a SETA has not complied with an instruction issued in terms of section 14A, the Minister may- (b) invoke section 15 without further notice to the SETA.” In other words, the power to invoke section 15 flows from section 14A. Reliance on section 15(4) applies only to bypass consultation, in 15(1), not to irrationally skip issuing instructions. The Act requires the Minister to issue an instruction before invoking section 15. This is basic Administrative Justice. Basic!

Sihle Lonzi

44,547 views • 10 months ago

🚨 xAI is moving fast right now. Grok 4.20 is being reported with major gains in reasoning, multimodal performance, instruction following, speed, and hallucination reduction, with its multi agent architecture. Currently #1 on IFBench at 82.9% Next best scores there are 79.6% and 79.0% Now Grok Imagine 1.0 is out of beta, and a bigger video release is already being teased for next week. xAI is shipping across reasoning, search, image, video, and agents all at once!

🚨 xAI is moving fast right now. Grok 4.20 is being reported with major gains in reasoning, multimodal performance, instruction following, speed, and hallucination reduction, with its multi agent architecture. Currently #1 on IFBench at 82.9% Next best scores there are 79.6% and 79.0% Now Grok Imagine 1.0 is out of beta, and a bigger video release is already being teased for next week. xAI is shipping across reasoning, search, image, video, and agents all at once!

tetsuo

18,369 views • 4 months ago

Super excited to introduce Pandora, a generative video World Model interactively controllable by language. #Sora and #GPT4 are both powerful. How about fusing them in a single model? 💥 Pandora gives a preview:🔭 > Build a General World Model (GWM) super efficiently by integrating pretrained autoregressive LLM and diffusion Video Model, aligning them in the representation space. > Let the LLM control the VM on-the-fly. Instruction tuning maximizes the controllability. > Autoregressive LLM empowers VM to generate indefinitely long videos: Starting with a VM for 2-second videos, Pandora extends it for 8-second videos. Would #Sora+#GPT4 under Pandora produce hours-long videos? 📽️ > World Model is beyond just video generation. It’s sensory-level information processing + concept-level reasoning and reflection. Pandora bridges both, with the concept- / language-level backbone (LLM) managing & steering the sensory-level VM functionalities. 👁️🧠 Check out for a bunch of interesting results:

Super excited to introduce Pandora, a generative video World Model interactively controllable by language. #Sora and #GPT4 are both powerful. How about fusing them in a single model? 💥 Pandora gives a preview:🔭 > Build a General World Model (GWM) super efficiently by integrating pretrained autoregressive LLM and diffusion Video Model, aligning them in the representation space. > Let the LLM control the VM on-the-fly. Instruction tuning maximizes the controllability. > Autoregressive LLM empowers VM to generate indefinitely long videos: Starting with a VM for 2-second videos, Pandora extends it for 8-second videos. Would #Sora+#GPT4 under Pandora produce hours-long videos? 📽️ > World Model is beyond just video generation. It’s sensory-level information processing + concept-level reasoning and reflection. Pandora bridges both, with the concept- / language-level backbone (LLM) managing & steering the sensory-level VM functionalities. 👁️🧠 Check out for a bunch of interesting results:

Zhiting Hu

63,456 views • 2 years ago

Rashid Latif urges Mohsin Naqvi to hold grand seminar with stakeholders and ex-cricketers to reform Pakistan cricket through collective input He also hinted at one notable name missing: “He’s not here…the one who should be invited is not here” A clear reference to Imran Khan.

Rashid Latif urges Mohsin Naqvi to hold grand seminar with stakeholders and ex-cricketers to reform Pakistan cricket through collective input He also hinted at one notable name missing: “He’s not here…the one who should be invited is not here” A clear reference to Imran Khan.

Pavilion Post

109,984 views • 2 months ago