Loading video...

Video Failed to Load

Go Home

Doc-to-LoRA: What if you could online distill documents into your LLM weights without training? 🚀 Stoked to share our new work on instant LLM adaptation using meta-learned hypernetworks 📷📝 Building on our previous Text-to-LoRA work, we doc-condition a hypernetwork to output LoRA adapters, improving the base LLM's effective context...

RobertTLange's profile picture

Robert Lange

10,000 subscribers

37,390 views • 7 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

🤔 Why do we still rely on the final layer of an LLM, when different layers encode different information? 🤔 In our new work, “Improving LLM Final Representations with Inter-Layer Geometry” (ICLR 2026 Workshop on Geometry-grounded Representation Learning and Generative Modeling) we show that actually, LLMs do not have one “best” layer. We introduce the Cayley-Encoder: an efficient and effective geometric encoder that learns one strong representation from all layer representations of the LLM, without biasing the representation toward any specific layer. While adding at most 0.1% learned parameters to the LLM, the Cayley-Encoder achieves large empirical gains over LoRA fine-tuning, final-layer representations, expensive attention-based aggregation, and methods that optimize specific layers for the task.
0:20

🤔 Why do we still rely on the final layer of an LLM, when different layers encode different information? 🤔 In our new work, “Improving LLM Final Representations with Inter-Layer Geometry” (ICLR 2026 Workshop on Geometry-grounded Representation Learning and Generative Modeling) we show that actually, LLMs do not have one “best” layer. We introduce the Cayley-Encoder: an efficient and effective geometric encoder that learns one strong representation from all layer representations of the LLM, without biasing the representation toward any specific layer. While adding at most 0.1% learned parameters to the LLM, the Cayley-Encoder achieves large empirical gains over LoRA fine-tuning, final-layer representations, expensive attention-based aggregation, and methods that optimize specific layers for the task.

Maya Bechler-Speicher

16,703 views • 3 months ago

Learn how to build an optimized LLM inference system from the ground up in our new short course, Efficiently Serving LLMs, built in collaboration with Predibase by Rubrik and taught by Travis Addair. Whether you're serving your own LLM or using a model hosting service, this course will give you a deep understanding of the optimizations required to efficiently serve many users at once. - Learn how LLMs generate text one token at a time, and how techniques like KV caching, continuous batching, and quantization speed things up and optimize memory usage for serving multiple users. - Benchmark the performance of these LLM optimizations to explore the trade-offs between quickly responding to an individual user’s request vs. serving many users at once. - Use techniques like low-rank adaptation (LoRA) to efficiently serve hundreds of unique, custom fine-tuned models on a single device, without sacrificing throughput. - Use Predibase's LoRAX framework to see optimization techniques in action on a real LLM server. Sign up here:
2:57

Learn how to build an optimized LLM inference system from the ground up in our new short course, Efficiently Serving LLMs, built in collaboration with Predibase by Rubrik and taught by Travis Addair. Whether you're serving your own LLM or using a model hosting service, this course will give you a deep understanding of the optimizations required to efficiently serve many users at once. - Learn how LLMs generate text one token at a time, and how techniques like KV caching, continuous batching, and quantization speed things up and optimize memory usage for serving multiple users. - Benchmark the performance of these LLM optimizations to explore the trade-offs between quickly responding to an individual user’s request vs. serving many users at once. - Use techniques like low-rank adaptation (LoRA) to efficiently serve hundreds of unique, custom fine-tuned models on a single device, without sacrificing throughput. - Use Predibase's LoRAX framework to see optimization techniques in action on a real LLM server. Sign up here:

Andrew Ng

104,727 views • 2 years ago

Learn a development pattern to systematically improve the accuracy and reliability of LLM applications in our new short course, Improving Accuracy of LLM Applications, built in partnership with Lamini and Meta, and taught by Lamini’s CEO Sharon Zhou, and Meta’s Senior Director of Partner Engineering, Amit Sangani. (Disclosure: I am an investor in Lamini.) The path to tuning an LLM application can be complex. In this course, you'll learn a systematic sequence of steps for improving accuracy by reducing hallucinations: - Create an evaluation dataset to measure model accuracy - Add prompt engineering and self-reflection - Fine-tune your model including "memory-tuning" which is a new method of embedding facts in an LLM Using the Llama 3-8B parameter model, you will: - Build a text-to-SQL agent with a custom schema and simulate situations where it hallucinates - Understand the difference between instruction fine-tuning, which gives pre-trained LLMs instructions to follow, and memory fine-tuning - See how Performance-Efficient Fine-tuning (PEFT) techniques like Low-Rank Adaptation (LoRA) reduce training time by 100x and Mixture of Memory Experts (MoME) reduces it even further I appreciate Meta releasing the Llama's family of open models -- this course gives an example of the unique type of work that developers can do with such models. Please sign up here:
3:18

Learn a development pattern to systematically improve the accuracy and reliability of LLM applications in our new short course, Improving Accuracy of LLM Applications, built in partnership with Lamini and Meta, and taught by Lamini’s CEO Sharon Zhou, and Meta’s Senior Director of Partner Engineering, Amit Sangani. (Disclosure: I am an investor in Lamini.) The path to tuning an LLM application can be complex. In this course, you'll learn a systematic sequence of steps for improving accuracy by reducing hallucinations: - Create an evaluation dataset to measure model accuracy - Add prompt engineering and self-reflection - Fine-tune your model including "memory-tuning" which is a new method of embedding facts in an LLM Using the Llama 3-8B parameter model, you will: - Build a text-to-SQL agent with a custom schema and simulate situations where it hallucinates - Understand the difference between instruction fine-tuning, which gives pre-trained LLMs instructions to follow, and memory fine-tuning - See how Performance-Efficient Fine-tuning (PEFT) techniques like Low-Rank Adaptation (LoRA) reduce training time by 100x and Mixture of Memory Experts (MoME) reduces it even further I appreciate Meta releasing the Llama's family of open models -- this course gives an example of the unique type of work that developers can do with such models. Please sign up here:

Andrew Ng

66,549 views • 2 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:
3:07

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

284,472 views • 2 years ago

An exciting new course: Fine-tuning and Reinforcement Learning for LLMs: Intro to Post-training, taught by Sharon Zhou, VP of AI at AMD. Available now at Post-training is the key technique used by frontier labs to turn a base LLM--a model trained on massive unlabeled text to predict the next word/token--into a helpful, reliable assistant that can follow instructions. I've also seen many applications where post-training is what turns a demo application that works only 80% of the time into a reliable system that consistently performs. This course will teach you the most important post-training techniques! In this 5 module course, Sharon walks you through the complete post-training pipeline: supervised fine-tuning, reward modeling, RLHF, and techniques like PPO and GRPO. You'll also learn to use LoRA for efficient training, and to design evals that catch problems before and after deployment. Skills you'll gain: - Apply supervised fine-tuning and reinforcement learning (RLHF, PPO, GRPO) to align models to desired behaviors - Use LoRA for efficient fine-tuning without retraining entire models - Prepare datasets and generate synthetic data for post-training - Understand how to operate LLM production pipelines, with go/no-go decision points and feedback loops These advanced methods aren’t limited to frontier AI labs anymore, and you can now use them in your own applications. Learn here:
2:27

An exciting new course: Fine-tuning and Reinforcement Learning for LLMs: Intro to Post-training, taught by Sharon Zhou, VP of AI at AMD. Available now at Post-training is the key technique used by frontier labs to turn a base LLM--a model trained on massive unlabeled text to predict the next word/token--into a helpful, reliable assistant that can follow instructions. I've also seen many applications where post-training is what turns a demo application that works only 80% of the time into a reliable system that consistently performs. This course will teach you the most important post-training techniques! In this 5 module course, Sharon walks you through the complete post-training pipeline: supervised fine-tuning, reward modeling, RLHF, and techniques like PPO and GRPO. You'll also learn to use LoRA for efficient training, and to design evals that catch problems before and after deployment. Skills you'll gain: - Apply supervised fine-tuning and reinforcement learning (RLHF, PPO, GRPO) to align models to desired behaviors - Use LoRA for efficient fine-tuning without retraining entire models - Prepare datasets and generate synthetic data for post-training - Understand how to operate LLM production pipelines, with go/no-go decision points and feedback loops These advanced methods aren’t limited to frontier AI labs anymore, and you can now use them in your own applications. Learn here:

Andrew Ng

133,386 views • 11 months ago

Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications. In this course, taught by Andreas Kollegger of Neo4j, you’ll - Explore how knowledge graphs work by building a graph of public financial documents from scratch - Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot - Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems Sign up here!
3:46

Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications. In this course, taught by Andreas Kollegger of Neo4j, you’ll - Explore how knowledge graphs work by building a graph of public financial documents from scratch - Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot - Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems Sign up here!

Andrew Ng

244,435 views • 2 years ago

Wrapping up the year and coinciding with #NeurIPS2024, today at Meta FAIR we’re releasing a collection of nine new open source AI research artifacts across our work in developing agents, robustness & safety and new architectures. More in the video from Joelle Pineau. All of this work is part of FAIR’s continued work towards the goal of achieving advanced machine intelligence A few highlights from what we’re releasing today: • Meta Motivo: A first-of-its-kind behavioral foundation model that controls the movements of a virtual embodied humanoid agent to perform complex tasks. • Meta Video Seal: a state-of-the art comprehensive framework for neural video watermarking. • Meta Explore Theory-of-Mind: A program-guided adversarial data generation for theory of mind reasoning. • Meta Large Concept Models: A fundamentally different training paradigm for language modeling that decouples reasoning from language representation. And much more! We’re excited to share this work with the research community and look forward to seeing how it inspires new innovation across the field. Details and access to everything released by FAIR today ➡️
1:34

Wrapping up the year and coinciding with #NeurIPS2024, today at Meta FAIR we’re releasing a collection of nine new open source AI research artifacts across our work in developing agents, robustness & safety and new architectures. More in the video from Joelle Pineau. All of this work is part of FAIR’s continued work towards the goal of achieving advanced machine intelligence A few highlights from what we’re releasing today: • Meta Motivo: A first-of-its-kind behavioral foundation model that controls the movements of a virtual embodied humanoid agent to perform complex tasks. • Meta Video Seal: a state-of-the art comprehensive framework for neural video watermarking. • Meta Explore Theory-of-Mind: A program-guided adversarial data generation for theory of mind reasoning. • Meta Large Concept Models: A fundamentally different training paradigm for language modeling that decouples reasoning from language representation. And much more! We’re excited to share this work with the research community and look forward to seeing how it inspires new innovation across the field. Details and access to everything released by FAIR today ➡️

AI at Meta

156,123 views • 1 year ago

Open science is how we continue to push technology forward and today at Meta FAIR we’re sharing eight new AI research artifacts including new models, datasets and code to inspire innovation in the community. More in the video from Joelle Pineau. This work is another important step towards our goal of achieving Advanced Machine Intelligence (AMI). What we’re releasing: • Meta Spirit LM: An open source language model for seamless speech and text integration. • Meta Segment Anything Model 2.1: An updated checkpoint with improved results on visually similar objects, small objects and occlusion handling. Plus a new developer suite to make it easier for developers to build with SAM 2. • Layer Skip: Inference code and fine-tuned checkpoints demonstrating a new method for enhancing LLM performance. • SALSA: New code to enable researchers to benchmark AI-based attacks in support of validating security for post-quantum cryptography. • Meta Lingua: A lightweight and self-contained codebase designed to train language models at scale. • Meta Open Materials: New open source models and the largest dataset of its kind to accelerate AI-driven discovery of new inorganic materials. • MEXMA: A new research paper and code for our novel pre-trained cross-lingual sentence encoder with coverage across 80 languages. • Self-Taught Evaluator: a new method for generating synthetic preference data to train reward models without relying on human annotations. Access to state-of-the-art AI creates opportunities for everyone. We’re excited to share this work and look forward to seeing the community innovation that results from it. Details and access to everything released by FAIR today ➡️
1:36

Open science is how we continue to push technology forward and today at Meta FAIR we’re sharing eight new AI research artifacts including new models, datasets and code to inspire innovation in the community. More in the video from Joelle Pineau. This work is another important step towards our goal of achieving Advanced Machine Intelligence (AMI). What we’re releasing: • Meta Spirit LM: An open source language model for seamless speech and text integration. • Meta Segment Anything Model 2.1: An updated checkpoint with improved results on visually similar objects, small objects and occlusion handling. Plus a new developer suite to make it easier for developers to build with SAM 2. • Layer Skip: Inference code and fine-tuned checkpoints demonstrating a new method for enhancing LLM performance. • SALSA: New code to enable researchers to benchmark AI-based attacks in support of validating security for post-quantum cryptography. • Meta Lingua: A lightweight and self-contained codebase designed to train language models at scale. • Meta Open Materials: New open source models and the largest dataset of its kind to accelerate AI-driven discovery of new inorganic materials. • MEXMA: A new research paper and code for our novel pre-trained cross-lingual sentence encoder with coverage across 80 languages. • Self-Taught Evaluator: a new method for generating synthetic preference data to train reward models without relying on human annotations. Access to state-of-the-art AI creates opportunities for everyone. We’re excited to share this work and look forward to seeing the community innovation that results from it. Details and access to everything released by FAIR today ➡️

AI at Meta

150,477 views • 1 year ago

As I promised yesterday, I'll briefly explain LoRA training and share a workflow I made so you can do it quickly. First, let me answer a very common question: 'Why train LoRAs when we have such advanced models?' Even though we have incredibly advanced models now (like NBP), we still can't always get them to do specific things we want. Simplest example: the spritesheet LoRA I made the other day. I generated 1000 images with Nano Banana and only 100 were what I wanted. The LoRA I trained using those 100 images gives me nearly 100% consistent results. Second point is cost and speed. With LoRA, we can cut costs by 4-5x. And while doing that, we're generating 4-5x faster. How many images do you need for a good LoRA? This depends on your LoRA's complexity. For example, when I training the spritesheet LoRA, even though I used 100 images, I didn't include buildings in the training data, so this LoRA doesn't work for buildings. So think about your LoRA's use cases and add examples for as many use cases as possible to improve quality. What are paired images and how to train LoRAs for image-editing? When training LoRAs for image editing on fal, we call each edit example paired images - one with _start suffix, one with _end suffix. For example, if you're training a background remove LoRA, the unedited original photo will be your '_start' image. The image with background removed will be the '_end' image. Simply put: images we want to edit or use as reference get _start, target images we want to achieve get '_end'. Important: save both images with the same name. Like image332_start.jpg and image332_end.jpg. This way the system knows which images pair together. What about training LoRAs for models with multiple image inputs? Same logic. We still use _start and _end suffixes, but with one difference. Since there are multiple input images, we can number them: _start, _start1, _start2. Example: start images, 1st image = Woman portrait (image35_start.jpg) 2nd image = Glasses photo (image35_start1.jpg) 3rd image = Hat photo (image35_start2.jpg) Output image = portrait of woman wearing glasses and hat (image35_end.jpg) Can we do more detailed captioning? Yes. Similarly, you can improve training quality by creating a txt file for each set with the caption inside. Example: create image35.txt and write: 'Recreate the image by putting the glasses from the second image and the hat from the third image on the woman in the first image.' What are Steps? How many should I use? What's Learning Rate? Steps determines how many times the model sees and processes your training data (your images). Each step, the model learns a bit more. But as steps increase, so does the risk of overfitting. So there's no real default. But for a simpler LoRA with 20 paired images, 1000 steps is ideal. Here's a metaphor for the Steps and Learning Rate relationship: Imagine you have a balloon. Our goal is to inflate it to the optimal size. Steps = How many times we blow into the balloon Learning rate = How hard we blow each time If we blow too softly, we need to blow many more times. If we blow too hard, we risk popping it quickly and can't reach optimal size. Of course training won't explode, but it won't work as intended because it wasn't trained optimally. Training's done, now what? Once training's complete, you'll have a safetensors file. Every model you train on fal has a LoRA inference endpoint. In that inference, add your safetensors file link to the LoRA url input, and you can use your LoRA. Thanks for the read! The workflow in the video: If I forgot anything, let me know in the replies.
1:06

As I promised yesterday, I'll briefly explain LoRA training and share a workflow I made so you can do it quickly. First, let me answer a very common question: 'Why train LoRAs when we have such advanced models?' Even though we have incredibly advanced models now (like NBP), we still can't always get them to do specific things we want. Simplest example: the spritesheet LoRA I made the other day. I generated 1000 images with Nano Banana and only 100 were what I wanted. The LoRA I trained using those 100 images gives me nearly 100% consistent results. Second point is cost and speed. With LoRA, we can cut costs by 4-5x. And while doing that, we're generating 4-5x faster. How many images do you need for a good LoRA? This depends on your LoRA's complexity. For example, when I training the spritesheet LoRA, even though I used 100 images, I didn't include buildings in the training data, so this LoRA doesn't work for buildings. So think about your LoRA's use cases and add examples for as many use cases as possible to improve quality. What are paired images and how to train LoRAs for image-editing? When training LoRAs for image editing on fal, we call each edit example paired images - one with _start suffix, one with _end suffix. For example, if you're training a background remove LoRA, the unedited original photo will be your '_start' image. The image with background removed will be the '_end' image. Simply put: images we want to edit or use as reference get _start, target images we want to achieve get '_end'. Important: save both images with the same name. Like image332_start.jpg and image332_end.jpg. This way the system knows which images pair together. What about training LoRAs for models with multiple image inputs? Same logic. We still use _start and _end suffixes, but with one difference. Since there are multiple input images, we can number them: _start, _start1, _start2. Example: start images, 1st image = Woman portrait (image35_start.jpg) 2nd image = Glasses photo (image35_start1.jpg) 3rd image = Hat photo (image35_start2.jpg) Output image = portrait of woman wearing glasses and hat (image35_end.jpg) Can we do more detailed captioning? Yes. Similarly, you can improve training quality by creating a txt file for each set with the caption inside. Example: create image35.txt and write: 'Recreate the image by putting the glasses from the second image and the hat from the third image on the woman in the first image.' What are Steps? How many should I use? What's Learning Rate? Steps determines how many times the model sees and processes your training data (your images). Each step, the model learns a bit more. But as steps increase, so does the risk of overfitting. So there's no real default. But for a simpler LoRA with 20 paired images, 1000 steps is ideal. Here's a metaphor for the Steps and Learning Rate relationship: Imagine you have a balloon. Our goal is to inflate it to the optimal size. Steps = How many times we blow into the balloon Learning rate = How hard we blow each time If we blow too softly, we need to blow many more times. If we blow too hard, we risk popping it quickly and can't reach optimal size. Of course training won't explode, but it won't work as intended because it wasn't trained optimally. Training's done, now what? Once training's complete, you'll have a safetensors file. Every model you train on fal has a LoRA inference endpoint. In that inference, add your safetensors file link to the LoRA url input, and you can use your LoRA. Thanks for the read! The workflow in the video: If I forgot anything, let me know in the replies.

ilker

15,192 views • 8 months ago

Mati Staniszewski on why ElevenLabs is betting on a cascaded approach for voice agents: “Our approach, as you think about voice agents, conversational agents, is effectively a cascaded approach. You use transcription or speech-to-text, an LLM, text-to-speech, and orchestrate all of that together. Then you have speech-to-speech, which goes directly from speech, and there’s a speech response on the other side. Today, we are optimizing heavily on a cascaded approach. As we work with a lot of the businesses and enterprises, they will need that visibility into what happens. They will want to execute certain tasks on top of that. They want good visibility into each of the steps and great accuracy of all the models. But beyond that, they can abstract away what’s the LLM layer, what’s the intelligence layer, and the integrations are easier in that system. That’s where we are betting a lot of the research work on how you can make that great, and we think we can make that great.” John Collison @matiii ElevenLabs
1:09

Mati Staniszewski on why ElevenLabs is betting on a cascaded approach for voice agents: “Our approach, as you think about voice agents, conversational agents, is effectively a cascaded approach. You use transcription or speech-to-text, an LLM, text-to-speech, and orchestrate all of that together. Then you have speech-to-speech, which goes directly from speech, and there’s a speech response on the other side. Today, we are optimizing heavily on a cascaded approach. As we work with a lot of the businesses and enterprises, they will need that visibility into what happens. They will want to execute certain tasks on top of that. They want good visibility into each of the steps and great accuracy of all the models. But beyond that, they can abstract away what’s the LLM layer, what’s the intelligence layer, and the integrations are easier in that system. That’s where we are betting a lot of the research work on how you can make that great, and we think we can make that great.” John Collison @matiii ElevenLabs

Stripe

18,605 views • 5 months ago

One great outcome of PaperWiki is personalized surveys. Survey papers continue to be one of the best ways to track a field. My agents are now generating personalized surveys on topics using my paper LLM-generated knowledge base. Self-improving and always up-to-date as new papers are published. Love how these topic wiki pages are evolving. One of my favorite views so far, but there is a lot more to do here. Plan to share these with our students at the DAIR.AI Academy so their agents can work on the frontier too. Stay tuned!
0:28

One great outcome of PaperWiki is personalized surveys. Survey papers continue to be one of the best ways to track a field. My agents are now generating personalized surveys on topics using my paper LLM-generated knowledge base. Self-improving and always up-to-date as new papers are published. Love how these topic wiki pages are evolving. One of my favorite views so far, but there is a lot more to do here. Plan to share these with our students at the DAIR.AI Academy so their agents can work on the frontier too. Stay tuned!

elvis

15,838 views • 5 months ago

Interesting approach to long context/continual learning here from Baseten at Cursor Compact long trajectories by compressing a prefix of the KV cache using an MLP/autoencoder. You can train this "compactor" MLP by learning to reconstruct activations that the original KV cache would produce on subsequent tokens. This maximally reconstructs information from long context that's useful for subsequent outputs. If you run an agent over extremely long context and run this compaction recursively, the activations of this compressed KV cache become like trained weights. Similar to task-specific LoRA or "cartridges". Would not be surprised if OpenAI is running a similar algorithm for their blackbox compaction. Clear benefits here if you can avoid busting the cache, as e.g. compacting via writing to a text file would. Seeing many emerging approaches to fixed size latent states for LLMs that seem promising. If building a task-specific KV cache compression ends up being more sample efficient than running backprop, getting this to "work" feels like one of the 2-3 remaining breakthroughs on the path to true AGI
13:13

Interesting approach to long context/continual learning here from Baseten at Cursor Compact long trajectories by compressing a prefix of the KV cache using an MLP/autoencoder. You can train this "compactor" MLP by learning to reconstruct activations that the original KV cache would produce on subsequent tokens. This maximally reconstructs information from long context that's useful for subsequent outputs. If you run an agent over extremely long context and run this compaction recursively, the activations of this compressed KV cache become like trained weights. Similar to task-specific LoRA or "cartridges". Would not be surprised if OpenAI is running a similar algorithm for their blackbox compaction. Clear benefits here if you can avoid busting the cache, as e.g. compacting via writing to a text file would. Seeing many emerging approaches to fixed size latent states for LLMs that seem promising. If building a task-specific KV cache compression ends up being more sample efficient than running backprop, getting this to "work" feels like one of the 2-3 remaining breakthroughs on the path to true AGI

Jay Hack

29,792 views • 3 months ago

Today we’re celebrating 10 years of the Meta FAIR lab in Paris by sharing a collection of new models, datasets and some exciting milestones in the impacts of open source — all laddering up to our ongoing work to achieve Advanced Machine Intelligence (AMI). 1️⃣ Meta PARTNR is a framework for human-robot collaboration that builds on our existing work in this space with a new dataset and a large planning model enabling robots to accomplish complex tasks alongside humans. 2️⃣ Meta Audiobox Aesthetics enables the automatic evaluation of audio aesthetics, providing a comprehensive assessment of audio quality across speech, music and sound. 3️⃣ Open Source Machine Translation Benchmark is a carefully crafted collection with the aim of building an unprecedented multilingual machine translation benchmark for the community. 4️⃣ Two new breakthrough studies using AI to further our understanding of language in the brain.
2:28

Today we’re celebrating 10 years of the Meta FAIR lab in Paris by sharing a collection of new models, datasets and some exciting milestones in the impacts of open source — all laddering up to our ongoing work to achieve Advanced Machine Intelligence (AMI). 1️⃣ Meta PARTNR is a framework for human-robot collaboration that builds on our existing work in this space with a new dataset and a large planning model enabling robots to accomplish complex tasks alongside humans. 2️⃣ Meta Audiobox Aesthetics enables the automatic evaluation of audio aesthetics, providing a comprehensive assessment of audio quality across speech, music and sound. 3️⃣ Open Source Machine Translation Benchmark is a carefully crafted collection with the aim of building an unprecedented multilingual machine translation benchmark for the community. 4️⃣ Two new breakthrough studies using AI to further our understanding of language in the brain.

AI at Meta

86,004 views • 1 year ago

ComfyUI Tutorial for Beginners: LoRAs, Style Transfer & ControlNets Module 3 is where you stop pulling the lever and start directing. Prompt iteration gets you close. These are the tools that get you exactly where you wanted to land. Style, structure, and pose, set before the sampler ever runs. What you'll learn: → How a LoRA attaches to a base model to change its output without retraining it → Why a LoRA has to match its base model family, and where to find them on Comfy Hub and Civitai → Model strength vs. CLIP strength — two dials, two different jobs → Stacking multiple LoRAs at different weights until the style is yours → Training your own: dataset, trigger word captions, step count, TrainLoRA locally → Reading the plot loss graph instead of guessing whether it worked → Appending to a LoRA you've already trained → Style transfer two ways: denoise-driven img2img vs. IP-Adapter reading style straight off a reference → ControlNets with DWPose, Depth Anything V2, and canny — composition locked before the prompt is read → Side by side: what each control map actually preserves → Fixing weak LoRAs, overcooked stacks, mismatched preprocessors, and noisy transfers Click the links below to run the workflow ⬇️
48:09

ComfyUI Tutorial for Beginners: LoRAs, Style Transfer & ControlNets Module 3 is where you stop pulling the lever and start directing. Prompt iteration gets you close. These are the tools that get you exactly where you wanted to land. Style, structure, and pose, set before the sampler ever runs. What you'll learn: → How a LoRA attaches to a base model to change its output without retraining it → Why a LoRA has to match its base model family, and where to find them on Comfy Hub and Civitai → Model strength vs. CLIP strength — two dials, two different jobs → Stacking multiple LoRAs at different weights until the style is yours → Training your own: dataset, trigger word captions, step count, TrainLoRA locally → Reading the plot loss graph instead of guessing whether it worked → Appending to a LoRA you've already trained → Style transfer two ways: denoise-driven img2img vs. IP-Adapter reading style straight off a reference → ControlNets with DWPose, Depth Anything V2, and canny — composition locked before the prompt is read → Side by side: what each control map actually preserves → Fixing weak LoRAs, overcooked stacks, mismatched preprocessors, and noisy transfers Click the links below to run the workflow ⬇️

ComfyUI

12,761 views • 22 days ago

Hi XR devs! Big day today. v83 of the Meta XR SDK is officially out and it includes something very special to me 🦙. For the past 6 months we’ve been building AI Inference Building Blocks, which you may have seen teased at Meta Connect 2025. Our goal: keep the simplicity of Building Blocks while giving you the freedom to plug in any AI provider or model you want to power you apps with AI. More details and a full step-by-step guide are coming soon. And stay tuned as two more iterations are already locked in for early next year with new capabilities and major performance and inference improvements like you've never seen on Meta Quest. 🚀 For now, learn more on the Meta Horizon OS Developer Center:
0:51

Hi XR devs! Big day today. v83 of the Meta XR SDK is officially out and it includes something very special to me 🦙. For the past 6 months we’ve been building AI Inference Building Blocks, which you may have seen teased at Meta Connect 2025. Our goal: keep the simplicity of Building Blocks while giving you the freedom to plug in any AI provider or model you want to power you apps with AI. More details and a full step-by-step guide are coming soon. And stay tuned as two more iterations are already locked in for early next year with new capabilities and major performance and inference improvements like you've never seen on Meta Quest. 🚀 For now, learn more on the Meta Horizon OS Developer Center:

Robi ᯅ

12,259 views • 9 months ago

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...
3:28

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...

kwindla

332,227 views • 1 month ago

Super excited to share 🧠MLGym 🦾 – the first Gym environment for AI Research Agents 🤖🔬 We introduce MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing LLM agents on AI research tasks. The key contributions of our work are: 🕹️ Enables the exploration of different training algorithms for AI Research Agents such as RL 🛠️ Provides a flexible evaluation framework that can accommodate different artifacts such as models, algorithms, or predictions 🤖 Allows researchers to evaluate any model without the need to develop a custom agentic harness 🎯 Introduces 13 diverse open-ended AI Research tasks for evaluating AI Research Agents on a wide range of domains such as computer vision, natural language processing, reinforcement learning, game theory, and logical reasoning. 📈 Proposes a new evaluation metric for AI Research Agents MLGym makes it easy to: 1) Add new tasks 2) Evaluate new models 3) Integrate new agents Check out a video of the MLGym Agent to see how it performs the full pipeline of idea generation💡, implementation 👩‍💻, experimentation 👩‍🔬, and iteration 🔄 to improve on ML tasks. Huge thanks to the exceptionally talented Deepak Nathani who led this work and to all the other amazing collaborators who made this possible 🙏🫶🚀
1:32

Super excited to share 🧠MLGym 🦾 – the first Gym environment for AI Research Agents 🤖🔬 We introduce MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing LLM agents on AI research tasks. The key contributions of our work are: 🕹️ Enables the exploration of different training algorithms for AI Research Agents such as RL 🛠️ Provides a flexible evaluation framework that can accommodate different artifacts such as models, algorithms, or predictions 🤖 Allows researchers to evaluate any model without the need to develop a custom agentic harness 🎯 Introduces 13 diverse open-ended AI Research tasks for evaluating AI Research Agents on a wide range of domains such as computer vision, natural language processing, reinforcement learning, game theory, and logical reasoning. 📈 Proposes a new evaluation metric for AI Research Agents MLGym makes it easy to: 1) Add new tasks 2) Evaluate new models 3) Integrate new agents Check out a video of the MLGym Agent to see how it performs the full pipeline of idea generation💡, implementation 👩‍💻, experimentation 👩‍🔬, and iteration 🔄 to improve on ML tasks. Huge thanks to the exceptionally talented Deepak Nathani who led this work and to all the other amazing collaborators who made this possible 🙏🫶🚀

Roberta Raileanu

105,157 views • 1 year ago

New Course: Post-training of LLMs Learn to post-train and customize an LLM in this short course, taught by Banghua Zhu, Assistant Professor at the University of Washington University of Washington, and co-founder of @NexusflowX. Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning. Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training. In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance. You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior. In detail, you’ll: - Understand what post-training is, when to use it, and how it differs from pre-training. - Build an SFT pipeline to turn a base model into an instruct model. - Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones. - Implement a DPO pipeline to change the identity of a chat assistant. - Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions. - Train a model with GRPO to improve its math capabilities using a verifiable reward. Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today. Please sign up here:
2:52

New Course: Post-training of LLMs Learn to post-train and customize an LLM in this short course, taught by Banghua Zhu, Assistant Professor at the University of Washington University of Washington, and co-founder of @NexusflowX. Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning. Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training. In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance. You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior. In detail, you’ll: - Understand what post-training is, when to use it, and how it differs from pre-training. - Build an SFT pipeline to turn a base model into an instruct model. - Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones. - Implement a DPO pipeline to change the identity of a chat assistant. - Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions. - Train a model with GRPO to improve its math capabilities using a verifiable reward. Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today. Please sign up here:

Andrew Ng

125,146 views • 1 year ago

fal's Gorkem Yurtseven on the Blender, Astra, and H3 Max workflow professionals are using to get exactly the shot they wanted: "Using Blender with one of these AI models together is an extremely popular workflow for professional work." "You render a low resolution of your scene, what you want to do, using Blender, previous non-AI technology. And once you add that video as a reference to an AI model, you basically get close to 100% controllability." "This is an incredibly popular workflow for VFX artists, people who are doing this professionally, because they want to get exactly what they put into the model." "A week after we launched H3 Max, people started generating these beautiful scenes using an LLM, Astra, in Blender, and all of a sudden it unlocked a whole new pipeline. Using an LLM to create a Blender scene and then passing that to the H3 Max model." "It works very well with H3 Max because it's extremely fast and you can try many things all at once in parallel." Gorkem Yurtseven batuhan the fal guy fal Jennifer Li
2:05

fal's Gorkem Yurtseven on the Blender, Astra, and H3 Max workflow professionals are using to get exactly the shot they wanted: "Using Blender with one of these AI models together is an extremely popular workflow for professional work." "You render a low resolution of your scene, what you want to do, using Blender, previous non-AI technology. And once you add that video as a reference to an AI model, you basically get close to 100% controllability." "This is an incredibly popular workflow for VFX artists, people who are doing this professionally, because they want to get exactly what they put into the model." "A week after we launched H3 Max, people started generating these beautiful scenes using an LLM, Astra, in Blender, and all of a sudden it unlocked a whole new pipeline. Using an LLM to create a Blender scene and then passing that to the H3 Max model." "It works very well with H3 Max because it's extremely fast and you can try many things all at once in parallel." Gorkem Yurtseven batuhan the fal guy fal Jennifer Li

a16z

34,883 views • 13 days ago