Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

1/ Six models, from 0.9B to 375B total parameters, released as one connected fleet. Institute of Foundation Models’s K2 Horizon is ambitious because it tries to make the training process inspectable, from pretraining to reasoning and agentic post-training. A closer look:

Chubby♨️

141,305 subscribers

22,026 просмотров • 1 месяц назад •via X (Twitter)

Наука и технологии Образование

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

Emerging as the global frontier of model-reasoning has been a core vision of Affine since its inception. In pursuit of this vision, we've been working relentlessly to forge a repeatable training loop: a mechanism that uses benchmark driven post-training methodology to facilitate the improvement of reasoning itself. Through applying our training methodology onto Qwen3-32B as our base model, Affine’s miners have been able to push the Pareto frontier outward. We’ve now reached saturation; our models have become so good at reasoning that any further improvements are producing diminishing returns. As a result we now move our base model to Qwen3.6-35B-A3B. It provides a stronger basis, offering an extended context window, as well as an improved foundation for agentic work. The same training methodology will be applied going forward. Our objective is no longer simply to improve a single model, but to demonstrate that the process in itself is repeatable, scalable, and model agnostic. We view the commoditization of reasoning as vital to ensure our technology is not only transparent, but also co-owned by humanity.

Emerging as the global frontier of model-reasoning has been a core vision of Affine since its inception. In pursuit of this vision, we've been working relentlessly to forge a repeatable training loop: a mechanism that uses benchmark driven post-training methodology to facilitate the improvement of reasoning itself. Through applying our training methodology onto Qwen3-32B as our base model, Affine’s miners have been able to push the Pareto frontier outward. We’ve now reached saturation; our models have become so good at reasoning that any further improvements are producing diminishing returns. As a result we now move our base model to Qwen3.6-35B-A3B. It provides a stronger basis, offering an extended context window, as well as an improved foundation for agentic work. The same training methodology will be applied going forward. Our objective is no longer simply to improve a single model, but to demonstrate that the process in itself is repeatable, scalable, and model agnostic. We view the commoditization of reasoning as vital to ensure our technology is not only transparent, but also co-owned by humanity.

Affine

14,866 просмотров • 4 месяцев назад

GLM 5.3’s biggest leap came from post-training alone. “GLM 5.3 was built entirely by post-training the previous model. For me, that was a wake-up call.” “Post-training is far more GPU-efficient than pre-training from scratch. Plenty of companies can’t afford to train a model from scratch, but they have proprietary data. They can post-train themselves and beat some of the closed models.” “There’s still so much juice left to squeeze from post-training. We compared GLM 5.3’s cyber capabilities with 5.2, and it’s a night and day difference. That’s all just post-training.”

GLM 5.3’s biggest leap came from post-training alone. “GLM 5.3 was built entirely by post-training the previous model. For me, that was a wake-up call.” “Post-training is far more GPU-efficient than pre-training from scratch. Plenty of companies can’t afford to train a model from scratch, but they have proprietary data. They can post-train themselves and beat some of the closed models.” “There’s still so much juice left to squeeze from post-training. We compared GLM 5.3’s cyber capabilities with 5.2, and it’s a night and day difference. That’s all just post-training.”

SemiAnalysis

214,839 просмотров • 20 дней назад

Fun conversation on No Priors w Liam Fedus on AI Foundation Model for Atoms from Periodic Labs -Foundation models that understand atoms -LLMs as reasoning engines -Timing to displace AI researchers Liam prior was VP Post Training at OpenAI, now CEO Periodic

Fun conversation on No Priors w Liam Fedus on AI Foundation Model for Atoms from Periodic Labs -Foundation models that understand atoms -LLMs as reasoning engines -Timing to displace AI researchers Liam prior was VP Post Training at OpenAI, now CEO Periodic

Elad Gil

25,995 просмотров • 7 месяцев назад

Today, we're joined by Aakanksha Chowdhery, member of technical staff at Reflection, to explore the fundamental shifts required to build true agentic AI. While the industry has largely focused on post-training techniques to improve reasoning, Aakanksha draws on her experience leading pre-training efforts for Google’s PaLM and early Gemini models to argue that pre-training itself must be rethought to move beyond static benchmarks. We explore the limitations of next-token prediction for multi-step workflows and examine how attention mechanisms, loss objectives, and training data must evolve to support long-form reasoning and planning. Aakanksha shares insights on the difference between context retrieval and actual reasoning, the importance of "trajectory" training data, and why scaling remains essential for discovering emergent agentic capabilities like error recovery and dynamic tool learning. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 02:26 - Reflection 04:54 - Limitations of post-training for building agents 07:31 - Rethinking pre-training in agents 10:51 - Scaling 11:27 - Evolving attention mechanisms for agentic capabilities 12:39 - Memory as a tool 14:13 - Loss objectives and training data 15:50 - Fine-tuning loss in agent performance 19:37 - Training data 21:29 - Augmenting dominant training data source 24:11 - Overcoming challenges in training on synthetic data 25:47 - Benchmarks 30:44 - Scaling laws in large models versus small models 33:20 - Long-form versus short-form reasoning 37:57 - Agent’s ability to recover from failure 40:15 - Hallucinations and failure recovery 43:53 - Tool use in agents 46:38 - Coding agents 48:37 - How researchers can contribute to agentic AI

Today, we're joined by Aakanksha Chowdhery, member of technical staff at Reflection, to explore the fundamental shifts required to build true agentic AI. While the industry has largely focused on post-training techniques to improve reasoning, Aakanksha draws on her experience leading pre-training efforts for Google’s PaLM and early Gemini models to argue that pre-training itself must be rethought to move beyond static benchmarks. We explore the limitations of next-token prediction for multi-step workflows and examine how attention mechanisms, loss objectives, and training data must evolve to support long-form reasoning and planning. Aakanksha shares insights on the difference between context retrieval and actual reasoning, the importance of "trajectory" training data, and why scaling remains essential for discovering emergent agentic capabilities like error recovery and dynamic tool learning. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 02:26 - Reflection 04:54 - Limitations of post-training for building agents 07:31 - Rethinking pre-training in agents 10:51 - Scaling 11:27 - Evolving attention mechanisms for agentic capabilities 12:39 - Memory as a tool 14:13 - Loss objectives and training data 15:50 - Fine-tuning loss in agent performance 19:37 - Training data 21:29 - Augmenting dominant training data source 24:11 - Overcoming challenges in training on synthetic data 25:47 - Benchmarks 30:44 - Scaling laws in large models versus small models 33:20 - Long-form versus short-form reasoning 37:57 - Agent’s ability to recover from failure 40:15 - Hallucinations and failure recovery 43:53 - Tool use in agents 46:38 - Coding agents 48:37 - How researchers can contribute to agentic AI

The TWIML AI Podcast

48,352 просмотров • 10 месяцев назад

Today we’re releasing K2 Think V2, our most capable open-source reasoning model to date. This is a fully sovereign model: trained end-to-end on IFM-curated and synthesized data, with complete transparency from pre-training through final reasoning alignment.

Today we’re releasing K2 Think V2, our most capable open-source reasoning model to date. This is a fully sovereign model: trained end-to-end on IFM-curated and synthesized data, with complete transparency from pre-training through final reasoning alignment.

MBZUAI

287,728 просмотров • 9 месяцев назад

Introducing Natura-1, our first general foundation model for the physical world. It follows natural-language instructions zero-shot and turns them into action with dexterous hands. Natura-1 is a 50-billion-parameter mixture-of-experts (MoE) model pretrained on one million hours of egocentric human experience. On top of this foundation, we conduct efficient post-training on a limited amount of robot data to adapt pretrained world knowledge to dexterous manipulation, yielding a model that generalizes to a wide range of unseen scenarios. Natura-1 follows complex natural-language instructions zero-shot, reasoning through spatial constraints, multi-object relationships, and negative conditions. It can also autonomously carry out long and continuous workflows to perform long-horizon executions. Natura-1 provides early evidence that generalization—and even emergent behavior—can coexist with the complexity of dexterous hands. Further progress depends on the full stack: infrastructure for faster training and real-time inference, pretraining data at greater scale and coverage, and models that transfer to robots with less adaptation. We expect it to be a long-term research and engineering effort.

Introducing Natura-1, our first general foundation model for the physical world. It follows natural-language instructions zero-shot and turns them into action with dexterous hands. Natura-1 is a 50-billion-parameter mixture-of-experts (MoE) model pretrained on one million hours of egocentric human experience. On top of this foundation, we conduct efficient post-training on a limited amount of robot data to adapt pretrained world knowledge to dexterous manipulation, yielding a model that generalizes to a wide range of unseen scenarios. Natura-1 follows complex natural-language instructions zero-shot, reasoning through spatial constraints, multi-object relationships, and negative conditions. It can also autonomously carry out long and continuous workflows to perform long-horizon executions. Natura-1 provides early evidence that generalization—and even emergent behavior—can coexist with the complexity of dexterous hands. Further progress depends on the full stack: infrastructure for faster training and real-time inference, pretraining data at greater scale and coverage, and models that transfer to robots with less adaptation. We expect it to be a long-term research and engineering effort.

nature will

1,043,474 просмотров • 25 дней назад

We are releasing Star Elastic - turn ONE reasoning LLM into MANY sizes with a single post-training run. 360× cheaper than pretraining a family of models. 7× better than SOTA compression. Split reasoning capability. Plus elastic budget control that beats the accuracy-latency frontier. Paper: HF models: Thread 👇

We are releasing Star Elastic - turn ONE reasoning LLM into MANY sizes with a single post-training run. 360× cheaper than pretraining a family of models. 7× better than SOTA compression. Split reasoning capability. Plus elastic budget control that beats the accuracy-latency frontier. Paper: HF models: Thread 👇

Pavlo Molchanov

22,304 просмотров • 5 месяцев назад

What makes an on-device agentic model actually useful? Post-training. Maxime Labonne, Edoardo Mosca, and Jiahui Wang from Liquid’s post-training team discuss how models learn to use tools, follow instructions, handle longer contexts, and recover when tasks get complex.

What makes an on-device agentic model actually useful? Post-training. Maxime Labonne, Edoardo Mosca, and Jiahui Wang from Liquid’s post-training team discuss how models learn to use tools, follow instructions, handle longer contexts, and recover when tasks get complex.

Liquid AI

42,403 просмотров • 1 месяц назад

Google's Jeff Dean says current pre-training is passive: initialize a model, stream the internet past it, let it observe But models need to learn not just from data, but by acting, predicting, and choosing what to learn from next "we have this artificial distinction now between pre and post-training, and it shouldn't exist long term"

Google's Jeff Dean says current pre-training is passive: initialize a model, stream the internet past it, let it observe But models need to learn not just from data, but by acting, predicting, and choosing what to learn from next "we have this artificial distinction now between pre and post-training, and it shouldn't exist long term"

Haider.

55,812 просмотров • 7 месяцев назад

🪄👩🏾‍💻🏗️ Simplify the process of pretraining and fine-tuning transformer models from scratch. This step-by-step guide breaks down how KerasNLP helps to make building state-of-the-art text processing models easier. ➡️

🪄👩🏾‍💻🏗️ Simplify the process of pretraining and fine-tuning transformer models from scratch. This step-by-step guide breaks down how KerasNLP helps to make building state-of-the-art text processing models easier. ➡️

TensorFlow

16,999 просмотров • 2 лет назад

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:

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

126,309 просмотров • 1 год назад

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote at #CES2026. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote at #CES2026. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️

NVIDIA AI Infrastructure

13,329 просмотров • 9 месяцев назад

FACT CHECK: No trainings OR standards have been cut from ICE training. New ICE recruits receieve 56 days of training and an average of 28 days of on-the-job training. Training increased from five days a week, eight hours a day to six days a week, twelve hours per day including personalized independent training. It’s the same hours of training officers have ALWAYS received.

FACT CHECK: No trainings OR standards have been cut from ICE training. New ICE recruits receieve 56 days of training and an average of 28 days of on-the-job training. Training increased from five days a week, eight hours a day to six days a week, twelve hours per day including personalized independent training. It’s the same hours of training officers have ALWAYS received.

Homeland Security

42,123 просмотров • 8 месяцев назад

Biggest takeaway from Day 1 of Eagles training camp? Jihaad Campbell can be an impact player right away — and the Eagles defense is going to be more dangerous because of it

Biggest takeaway from Day 1 of Eagles training camp? Jihaad Campbell can be an impact player right away — and the Eagles defense is going to be more dangerous because of it

Eliot Shorr-Parks

56,626 просмотров • 1 год назад

“We have a crisis of open source models in the Western world. Outside of China, there are no good open source models. We don't even have any in the US now. The talent, the capital and the focus to be best-in-class at pre-training, mid-training, post-training is an extremely scarce skillset. The answer for a lot of countries may just be to take an open weights model from China, post-train or fine-tune their own version, and have that be what they start from." Everett Randle Why does the West have such poor open source models and is it a national security threat to rely on open-source Chinese models Demis Hassabis Yann LeCun Aidan Gomez Lilian Weng Jan Leike

“We have a crisis of open source models in the Western world. Outside of China, there are no good open source models. We don't even have any in the US now. The talent, the capital and the focus to be best-in-class at pre-training, mid-training, post-training is an extremely scarce skillset. The answer for a lot of countries may just be to take an open weights model from China, post-train or fine-tune their own version, and have that be what they start from." Everett Randle Why does the West have such poor open source models and is it a national security threat to rely on open-source Chinese models Demis Hassabis Yann LeCun Aidan Gomez Lilian Weng Jan Leike

Harry Stebbings

63,738 просмотров • 4 месяцев назад

Natural energy. Barefoot on the Earth, connected to the Source. Real power doesn’t make noise — it flows from the ground up through hard training and deep breathing. A connection that can’t be broken.

Natural energy. Barefoot on the Earth, connected to the Source. Real power doesn’t make noise — it flows from the ground up through hard training and deep breathing. A connection that can’t be broken.

McGregor News

10,729 просмотров • 6 месяцев назад

Join the Tooling, Process and Training Team to help develop the equipment that will help build a world of amazing abundance – Design advanced tooling – Streamline service processes – Deliver new training approaches to maximize fleet health Roles based in Hayward, California

Join the Tooling, Process and Training Team to help develop the equipment that will help build a world of amazing abundance – Design advanced tooling – Streamline service processes – Deliver new training approaches to maximize fleet health Roles based in Hayward, California

Tesla Recruiting

60,595 просмотров • 8 месяцев назад

Evolution happens to populations. Useful genes become common in a group over generations, then spread to other groups that test them in their own environments. The speed depends on generation time, 20 minutes for a bacterium but about 25 years for an elephant. Using LLMs to evolve better models often looks like an isolated herd of elephants. Each search runs alone on one dataset, and for large models every generation costs a long training run on the full data. Our new paper, "Discovered, Not Designed," introduces Population Evolution. It has two goals, to discover better models and to discover them faster. To discover better model designs, the searches share what they find. Several searches run at the same time, each on its own dataset. When one finds a code change that helps, the same change is retrained on the others. While nature keeps no record of the genes that failed to help, Population Evolution does. All peer results, including weak or unsuccessful transfers, are added to a shared ledger, and the LLM reads that ledger before it writes the next change. To discover them faster, the searches do not train on the full data, because that is what makes each generation slow. We found that a search can run on as little as 2% or 5% of the data and still find changes that hold up in full training, as long as the other searches check each change first. In Population Evolution, a change that works on one slice must also work on at least two other slices, then on two larger 10% sets. Only then does it get a full training run. We tested Population Evolution on RMD-Bench, a benchmark we also introduce for evaluating model discovery. It covers five task families (LLM pretraining, VLM pretraining, RL algorithm discovery, ranking, and watch-time prediction) with 34 instances in total. Each one comes with a reference implementation and a fixed training and evaluation protocol, so a change found on one dataset or data scale can be retrained and scored on another, and other discovery methods can be compared under the same conditions. The numbers below are a sneak peek at the two most expensive families of experiments. The LLM pretraining task starts from a 185M-parameter Gemma 3-style model, and the VLM pretraining task starts from a Qwen3.5-style model with about 646M trainable parameters. The baseline runs the same kind of search directly on the full data, at the same total GPU compute. Population Evolution finds better models. The best relative accuracy gain for LLM pretraining goes from 0.92% to 2.48%. The best loss reduction for VLM pretraining goes from 5.05% to 8.78%. Population Evolution finds them faster. In LLM pretraining, each search gets 100 generations where a full-data search gets 14. In VLM pretraining, 57 of 61 full training runs met the success threshold, against 12 of 100. Evolution works because many groups pass along what helps and get many generations to do it, and Population Evolution gives LLM-driven model search these prerequisites. Read our paper at for full implementation details and results. The benchmark data and code will be made publicly available shortly.

Evolution happens to populations. Useful genes become common in a group over generations, then spread to other groups that test them in their own environments. The speed depends on generation time, 20 minutes for a bacterium but about 25 years for an elephant. Using LLMs to evolve better models often looks like an isolated herd of elephants. Each search runs alone on one dataset, and for large models every generation costs a long training run on the full data. Our new paper, "Discovered, Not Designed," introduces Population Evolution. It has two goals, to discover better models and to discover them faster. To discover better model designs, the searches share what they find. Several searches run at the same time, each on its own dataset. When one finds a code change that helps, the same change is retrained on the others. While nature keeps no record of the genes that failed to help, Population Evolution does. All peer results, including weak or unsuccessful transfers, are added to a shared ledger, and the LLM reads that ledger before it writes the next change. To discover them faster, the searches do not train on the full data, because that is what makes each generation slow. We found that a search can run on as little as 2% or 5% of the data and still find changes that hold up in full training, as long as the other searches check each change first. In Population Evolution, a change that works on one slice must also work on at least two other slices, then on two larger 10% sets. Only then does it get a full training run. We tested Population Evolution on RMD-Bench, a benchmark we also introduce for evaluating model discovery. It covers five task families (LLM pretraining, VLM pretraining, RL algorithm discovery, ranking, and watch-time prediction) with 34 instances in total. Each one comes with a reference implementation and a fixed training and evaluation protocol, so a change found on one dataset or data scale can be retrained and scored on another, and other discovery methods can be compared under the same conditions. The numbers below are a sneak peek at the two most expensive families of experiments. The LLM pretraining task starts from a 185M-parameter Gemma 3-style model, and the VLM pretraining task starts from a Qwen3.5-style model with about 646M trainable parameters. The baseline runs the same kind of search directly on the full data, at the same total GPU compute. Population Evolution finds better models. The best relative accuracy gain for LLM pretraining goes from 0.92% to 2.48%. The best loss reduction for VLM pretraining goes from 5.05% to 8.78%. Population Evolution finds them faster. In LLM pretraining, each search gets 100 generations where a full-data search gets 14. In VLM pretraining, 57 of 61 full training runs met the success threshold, against 12 of 100. Evolution works because many groups pass along what helps and get many generations to do it, and Population Evolution gives LLM-driven model search these prerequisites. Read our paper at for full implementation details and results. The benchmark data and code will be made publicly available shortly.

Zhuokai Zhao

925,744 просмотров • 1 месяц назад