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.

Introducing Flow Reasoning Models. We developed a recurrent flow-based architecture to efficiently solve structured reasoning problems (e.g., Sudoku). FRMs apply continuous flows to discrete data and recurrently refine their past mistakes through self-conditioning.

Alec Helbling

11,351 subscribers

103,781 views • 1 day ago •via X (Twitter)

Science & Technology Education

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

GPT-5 on Sudoku-Bench 🧩 Since releasing Sudoku-Bench in May 2025, when no LLM could solve a classic 9x9 puzzle, we've been evaluating the latest generation of models. GPT-5 now leads our leaderboard with 33% puzzles solved--approximately 2x the previous leader--and is the first LLM we've tested to solve a 9x9 Sudoku variant. However, with 67% of the much harder puzzles remaining unsolved, Sudoku-Bench continues to present significant challenges for AI reasoning. Modern Sudoku variants require models to first understand novel rulesets through meta-reasoning, then maintain global consistency across long reasoning chains. Our experiments with GRPO fine-tuning on Qwen2.5-7b and "Thought Cloning" (training on expert human reasoning from Cracking the Cryptic) show that current approaches still struggle with the spatial reasoning and creative "break-in" points that human solvers use naturally. We believe new approaches are required to solve our benchmark. These results highlight persistent gaps between computational problem-solving and human-like reasoning, particularly in tasks requiring integrated mathematical logic, spatial awareness, and creative insight. Read more about our update here: 🔗 Blogpost →

GPT-5 on Sudoku-Bench 🧩 Since releasing Sudoku-Bench in May 2025, when no LLM could solve a classic 9x9 puzzle, we've been evaluating the latest generation of models. GPT-5 now leads our leaderboard with 33% puzzles solved--approximately 2x the previous leader--and is the first LLM we've tested to solve a 9x9 Sudoku variant. However, with 67% of the much harder puzzles remaining unsolved, Sudoku-Bench continues to present significant challenges for AI reasoning. Modern Sudoku variants require models to first understand novel rulesets through meta-reasoning, then maintain global consistency across long reasoning chains. Our experiments with GRPO fine-tuning on Qwen2.5-7b and "Thought Cloning" (training on expert human reasoning from Cracking the Cryptic) show that current approaches still struggle with the spatial reasoning and creative "break-in" points that human solvers use naturally. We believe new approaches are required to solve our benchmark. These results highlight persistent gaps between computational problem-solving and human-like reasoning, particularly in tasks requiring integrated mathematical logic, spatial awareness, and creative insight. Read more about our update here: 🔗 Blogpost →

Sakana AI

154,872 views • 9 months ago

Chain-of-thought reasoning is a powerful tool to enable language models to work through complex problems. Can we use this with robots? With embodied chain-of-thought, vision-language-action (VLA) models can think through perception and planning! A 🧵👇

Chain-of-thought reasoning is a powerful tool to enable language models to work through complex problems. Can we use this with robots? With embodied chain-of-thought, vision-language-action (VLA) models can think through perception and planning! A 🧵👇

Sergey Levine

30,479 views • 2 years ago

We just brought flow maps to language modeling for one-step sequence generation 💥 Discrete diffusion is not necessary -- continuous flows over one-hot encodings achieve SoTA performance and ≥8.3× faster generation 🔥 We believe this is a major step forward for discrete generative modeling and language modeling alike. 🚀 Full thread from first author Chanhyuk Lee:

We just brought flow maps to language modeling for one-step sequence generation 💥 Discrete diffusion is not necessary -- continuous flows over one-hot encodings achieve SoTA performance and ≥8.3× faster generation 🔥 We believe this is a major step forward for discrete generative modeling and language modeling alike. 🚀 Full thread from first author Chanhyuk Lee:

Nicholas Boffi

44,442 views • 6 months ago

💡Divergence thinking💡 is a hallmark of human creativity and problem-solving 🤖Can LLMs also do divergent reasoning to generate diverse solutions🤔? Introducing Flow-of-Reasoning (FoR) 🌊, a data-efficient way of training LLM policy to generate diverse, high-quality reasoning trajectories Unlike existing RL (like PPO) and planning (like MCTS) to find the max-reward trajectory (akin to convergent thinking), FoR connects LLM reasoning with the #GFlowNet formulation and enables LLMs to find trajectories proportional to reward distribution. 🎬The demo video illustrates how FoR learns and infers multiple solutions to a ♠️Game24 puzzle. 🎯Inferring for diverse solutions could be useful for robustness, data augmentation, and enhanced model generalization. Project page: Paper: Github:

💡Divergence thinking💡 is a hallmark of human creativity and problem-solving 🤖Can LLMs also do divergent reasoning to generate diverse solutions🤔? Introducing Flow-of-Reasoning (FoR) 🌊, a data-efficient way of training LLM policy to generate diverse, high-quality reasoning trajectories Unlike existing RL (like PPO) and planning (like MCTS) to find the max-reward trajectory (akin to convergent thinking), FoR connects LLM reasoning with the #GFlowNet formulation and enables LLMs to find trajectories proportional to reward distribution. 🎬The demo video illustrates how FoR learns and infers multiple solutions to a ♠️Game24 puzzle. 🎯Inferring for diverse solutions could be useful for robustness, data augmentation, and enhanced model generalization. Project page: Paper: Github:

Lianhui Qin

50,493 views • 2 years ago

#AGIBOTAIWeek Day 3: Introducing GO-2: the next-generation foundation model for embodied AI, built to unify reasoning and action. To truly bridge “thinking” and “doing,” embodied AI must solve two challenges at once: • It generate executable action plans through deep spatial reasoning • It deliver stable execution in real-world environments GO-2 tackles both with a comprehensive architecture: Action Chain-of-Thought for action reasoning, and an Asynchronous Dual-System for robust execution. #AGIBOT #AGIBOTAIWeek #Foundation #model #EmbodiedAI Official account tags X: AGIBOT LinkedIn: agibot

#AGIBOTAIWeek Day 3: Introducing GO-2: the next-generation foundation model for embodied AI, built to unify reasoning and action. To truly bridge “thinking” and “doing,” embodied AI must solve two challenges at once: • It generate executable action plans through deep spatial reasoning • It deliver stable execution in real-world environments GO-2 tackles both with a comprehensive architecture: Action Chain-of-Thought for action reasoning, and an Asynchronous Dual-System for robust execution. #AGIBOT #AGIBOTAIWeek #Foundation #model #EmbodiedAI Official account tags X: AGIBOT LinkedIn: agibot

Hasan Toor

36,202 views • 4 months ago

.Trenton Bricken explains how we know LLMs are actually generalizing - aka they're not just stochastic parrots: - Training models on code makes them better at reasoning in language. - Models fine tuned on math problems become better at entity detection. - We can just straightforwardly read the world-models developed by smaller NNs which are easier to interpret (Othello). Transfer learning shows models are developing a deeper understanding of their data. Full episode out Thursday.

.Trenton Bricken explains how we know LLMs are actually generalizing - aka they're not just stochastic parrots: - Training models on code makes them better at reasoning in language. - Models fine tuned on math problems become better at entity detection. - We can just straightforwardly read the world-models developed by smaller NNs which are easier to interpret (Othello). Transfer learning shows models are developing a deeper understanding of their data. Full episode out Thursday.

Dwarkesh Patel

237,812 views • 2 years ago

Building AI agents? 🚧 Make sure they actually know where their answers come from. As Brana Rakic demonstrates, scalable AI requires verifiable knowledge, rule-based reasoning, and LLMs grounded in trusted memory. Key highlights: 03:25 - From human data to era of experience 07:42 - Agent architecture: memory + models 15:18 - Knowledge graphs 101: triples explained 34:53 - Symbolic reasoning vs hallucination 39:03 - Knowledge Asset, UAL & ownership

Building AI agents? 🚧 Make sure they actually know where their answers come from. As Brana Rakic demonstrates, scalable AI requires verifiable knowledge, rule-based reasoning, and LLMs grounded in trusted memory. Key highlights: 03:25 - From human data to era of experience 07:42 - Agent architecture: memory + models 15:18 - Knowledge graphs 101: triples explained 34:53 - Symbolic reasoning vs hallucination 39:03 - Knowledge Asset, UAL & ownership

OriginTrail

17,750 views • 6 months ago

Geoffrey Hinton says the current path of scaling is hitting a limit Most high-value data is locked inside companies, and the "free internet" is largely exhausted The solution is for models to generate their own training data through reasoning "that's how AlphaGo beat humans"

Geoffrey Hinton says the current path of scaling is hitting a limit Most high-value data is locked inside companies, and the "free internet" is largely exhausted The solution is for models to generate their own training data through reasoning "that's how AlphaGo beat humans"

Haider.

149,442 views • 8 months ago

Humans learn and improve from failures. Similarly, foundation models adapt based on human feedback. Can we leverage this failure understanding to enhance robotics systems that use foundation models? Introducing AHA—a vision-language model for detecting and reasoning over failures in robotic manipulation. Project page: 🧵Thread👇 Aha!

Humans learn and improve from failures. Similarly, foundation models adapt based on human feedback. Can we leverage this failure understanding to enhance robotics systems that use foundation models? Introducing AHA—a vision-language model for detecting and reasoning over failures in robotic manipulation. Project page: 🧵Thread👇 Aha!

Jiafei Duan

48,827 views • 1 year ago

Current world models often separate spatial reasoning from temporal modeling, leading to consistency bottlenecks and scaling limits. We developed RAYNOVA, a 4D world foundation model, to address this by unifying space and time into a single representation via a pure auto-regressive framework. The result: Data-driven high-fidelity simulation data that scales across any vehicles, anywhere. Read the full technical breakdown:

Current world models often separate spatial reasoning from temporal modeling, leading to consistency bottlenecks and scaling limits. We developed RAYNOVA, a 4D world foundation model, to address this by unifying space and time into a single representation via a pure auto-regressive framework. The result: Data-driven high-fidelity simulation data that scales across any vehicles, anywhere. Read the full technical breakdown:

Applied Intuition

11,252 views • 3 months ago

Introducing Reinforcement-Learned Teachers (RLTs): Transforming how we teach LLMs to reason with reinforcement learning (RL). Blog: Paper: Traditional RL focuses on “learning to solve” challenging problems with expensive LLMs and constitutes a key step in making student AI systems ultimately acquire reasoning capabilities via distillation and cold-starting. Enter our RLTs—a new class of models prompted with not only a problem’s question but also its solution, and directly trained to generate clear, step-by-step “explanations” to teach their students. Remarkably, an RLT with only 7B parameters produces superior results when distilling and cold-starting students in competitive and graduate-level reasoning tasks than orders-of-magnitude larger LLMs. RLTs are as effective even when distilling 32B students, much larger than the teacher itself—unlocking a new standard for efficiency in developing reasoning language models with RL. Code:

Introducing Reinforcement-Learned Teachers (RLTs): Transforming how we teach LLMs to reason with reinforcement learning (RL). Blog: Paper: Traditional RL focuses on “learning to solve” challenging problems with expensive LLMs and constitutes a key step in making student AI systems ultimately acquire reasoning capabilities via distillation and cold-starting. Enter our RLTs—a new class of models prompted with not only a problem’s question but also its solution, and directly trained to generate clear, step-by-step “explanations” to teach their students. Remarkably, an RLT with only 7B parameters produces superior results when distilling and cold-starting students in competitive and graduate-level reasoning tasks than orders-of-magnitude larger LLMs. RLTs are as effective even when distilling 32B students, much larger than the teacher itself—unlocking a new standard for efficiency in developing reasoning language models with RL. Code:

Sakana AI

179,362 views • 1 year ago

Today we are introducing Tara. Biological datasets are a source of insights and a means to train biological AI models. As the ability to reason at scale emerges, they take on a new role: the ground truth for testing what reasoning models produce, and the environment in which those models operate, get feedback, and improve. Tara, our autonomous research agent, is embedded in our ever-expanding datasets, lab-generated and synthetic, and built to test and evolve the hypotheses frontier models generate, matching the pace at which they produce new ideas. By keeping those models grounded in a vast space of high-precision biological data, we believe we can compound biological reasoning and close the impedance mismatch between hypothesis generation and validation.

Today we are introducing Tara. Biological datasets are a source of insights and a means to train biological AI models. As the ability to reason at scale emerges, they take on a new role: the ground truth for testing what reasoning models produce, and the environment in which those models operate, get feedback, and improve. Tara, our autonomous research agent, is embedded in our ever-expanding datasets, lab-generated and synthetic, and built to test and evolve the hypotheses frontier models generate, matching the pace at which they produce new ideas. By keeping those models grounded in a vast space of high-precision biological data, we believe we can compound biological reasoning and close the impedance mismatch between hypothesis generation and validation.

Nima Alidoust

30,365 views • 2 months ago

Jensen Huang explains that AI has progressed through four major phases. The first wave was Perception AI, where deep learning enabled superhuman vision and speech recognition. The second wave was Generative AI, allowing AI to create text, images, video, and more. We are now in the third wave, Reasoning AI, where models can think through problems, apply logic, and even conduct research. The fourth wave will be Physical AI, where models gain real-world understanding and common sense, leading to advanced robotics.

Jensen Huang explains that AI has progressed through four major phases. The first wave was Perception AI, where deep learning enabled superhuman vision and speech recognition. The second wave was Generative AI, allowing AI to create text, images, video, and more. We are now in the third wave, Reasoning AI, where models can think through problems, apply logic, and even conduct research. The fourth wave will be Physical AI, where models gain real-world understanding and common sense, leading to advanced robotics.

Wes Roth

69,406 views • 1 year ago

Check out our #PAMI paper with code "Dense Continuous-Time Optical Flow from Event Cameras," where we show how to regress *continuous-time* trajectories of every pixel from event cameras alone or events plus frames! The key idea is to iteratively estimate per-pixel polynomials using a recurrent lookup and update scheme. Paper: Code: DOI: We present a method for estimating dense continuous-time optical flow from event data. Traditional dense optical flow methods compute the pixel displacement between two images. Due to missing information, these approaches cannot recover the pixel trajectories in the blind time between two images. We show that it is possible to compute per-pixel, continuous-time optical flow using events from an event camera. Events provide temporally fine-grained information about movement in pixel space due to their asynchronous nature and microsecond response time. We leverage these benefits to predict pixel trajectories densely in continuous time via parameterized Bézier curves. To achieve this, we build a neural network with strong inductive biases for this task: First, we build multiple sequential correlation volumes in time using event data. Second, we use Bézier curves to index these correlation volumes at multiple timestamps along the trajectory. Third, we use the retrieved correlation to update the Bézier curve representations iteratively. Our method can optionally include image pairs to boost performance further. To train and evaluate our model, we introduce a synthetic dataset (MultiFlow) that features moving objects and ground truth trajectories for every pixel. Our quantitative experiments suggest that our method successfully predicts pixel trajectories in continuous time and is competitive in the traditional two-view pixel displacement metric on MultiFlow and DSEC-Flow. Open source code and datasets are released to the public. Kudos to Mathias Gehrig Manasi Muglikar

Check out our #PAMI paper with code "Dense Continuous-Time Optical Flow from Event Cameras," where we show how to regress continuous-time trajectories of every pixel from event cameras alone or events plus frames! The key idea is to iteratively estimate per-pixel polynomials using a recurrent lookup and update scheme. Paper: Code: DOI: We present a method for estimating dense continuous-time optical flow from event data. Traditional dense optical flow methods compute the pixel displacement between two images. Due to missing information, these approaches cannot recover the pixel trajectories in the blind time between two images. We show that it is possible to compute per-pixel, continuous-time optical flow using events from an event camera. Events provide temporally fine-grained information about movement in pixel space due to their asynchronous nature and microsecond response time. We leverage these benefits to predict pixel trajectories densely in continuous time via parameterized Bézier curves. To achieve this, we build a neural network with strong inductive biases for this task: First, we build multiple sequential correlation volumes in time using event data. Second, we use Bézier curves to index these correlation volumes at multiple timestamps along the trajectory. Third, we use the retrieved correlation to update the Bézier curve representations iteratively. Our method can optionally include image pairs to boost performance further. To train and evaluate our model, we introduce a synthetic dataset (MultiFlow) that features moving objects and ground truth trajectories for every pixel. Our quantitative experiments suggest that our method successfully predicts pixel trajectories in continuous time and is competitive in the traditional two-view pixel displacement metric on MultiFlow and DSEC-Flow. Open source code and datasets are released to the public. Kudos to Mathias Gehrig Manasi Muglikar

Davide Scaramuzza

12,681 views • 2 years ago

we achieved some progress in the implementation of state inference inside octra , first we managed to implement continuous reasoning without clearing the circle memory, second we completed a pure resource model under which validators can verify and certify the execution flow of a set of useful work (in this case this is ML inference with text generation by input tokens fully onchain), a few months ago the answers of the test model were one-word and substantially limited, in the current version (which will be available for everyone) the reasoning flow is automatically extended, it will be possible to upload various models into circles or create tools for their use making them public or only private (full demo will be opensource as soon as we finish, it will be possible to deploy anything in 2-3 clicks) live use case, the economic model also converges here - wallet pays for inference in native coins, validators (with GPU support can maintain availability of their resources to quickly process such private calls, while the execution core remains private through HFHE)

we achieved some progress in the implementation of state inference inside octra , first we managed to implement continuous reasoning without clearing the circle memory, second we completed a pure resource model under which validators can verify and certify the execution flow of a set of useful work (in this case this is ML inference with text generation by input tokens fully onchain), a few months ago the answers of the test model were one-word and substantially limited, in the current version (which will be available for everyone) the reasoning flow is automatically extended, it will be possible to upload various models into circles or create tools for their use making them public or only private (full demo will be opensource as soon as we finish, it will be possible to deploy anything in 2-3 clicks) live use case, the economic model also converges here - wallet pays for inference in native coins, validators (with GPU support can maintain availability of their resources to quickly process such private calls, while the execution core remains private through HFHE)

λ

11,482 views • 29 days ago

Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents. Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold. Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.

Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents. Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold. Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.

OpenAI

3,663,601 views • 4 months ago

Introducing Nightglass — the options flow platform built to filter out the noise and show you the 2% of trades worth noticing. Hedge funds and institutions place bets through options every day, and their trades are public. You just have to know how to spot them. The problem: 98% of options flow is noise. That's why we built Nightglass — it doesn't show you more flow, it shows you less. Then it gives you the plan to trade it. -Institutional support and resistance levels on any US-listed stock -Real-time flow alerts -Pre-market report of the prior session's institutional positioning -Live SPX gamma exposure levels -Earnings setups with defined trade structures -A morning brief before every open with key news, sell side research, top flows, and more. Live today.

Introducing Nightglass — the options flow platform built to filter out the noise and show you the 2% of trades worth noticing. Hedge funds and institutions place bets through options every day, and their trades are public. You just have to know how to spot them. The problem: 98% of options flow is noise. That's why we built Nightglass — it doesn't show you more flow, it shows you less. Then it gives you the plan to trade it. -Institutional support and resistance levels on any US-listed stock -Real-time flow alerts -Pre-market report of the prior session's institutional positioning -Live SPX gamma exposure levels -Earnings setups with defined trade structures -A morning brief before every open with key news, sell side research, top flows, and more. Live today.

Nightglass

14,470 views • 18 days ago

With enough data, robots and AI can learn “world models” that let them predict the results of their actions. These models are a way to learn how embodied AI agents can perform a wide variety of useful tasks — but they require a huge amount of data. The team at General Intuition General Intuition has a solution: use data from video games! Games teach movement, problem solving, and complex spatial reasoning, and they come in a staggering diversity of forms, covering a wide variety of problems. What’s more, the captured data is high-quality, without the noise or annotation error that can come from We sat down with Pim de Witte and Adam Jelley from the General Intuition team to learn more about their history, their plans, and their philosophy.

With enough data, robots and AI can learn “world models” that let them predict the results of their actions. These models are a way to learn how embodied AI agents can perform a wide variety of useful tasks — but they require a huge amount of data. The team at General Intuition General Intuition has a solution: use data from video games! Games teach movement, problem solving, and complex spatial reasoning, and they come in a staggering diversity of forms, covering a wide variety of problems. What’s more, the captured data is high-quality, without the noise or annotation error that can come from We sat down with Pim de Witte and Adam Jelley from the General Intuition team to learn more about their history, their plans, and their philosophy.

RoboPapers

85,927 views • 9 months ago

NEWS: NVIDIA just announced Alpamayo, what CEO Jensen Huang calls the world’s first thinking, reasoning autonomous vehicle AI, launching on U.S. roads later this year, starting with the Mercedes CLA. Jensen: "It's trained end-to-end. Literally from camera in to actuation out; It reasons what action it is about to take, the reason by which is came about that action, and the trajectory." Alpamayo introduces Vision-Language-Action (VLA) models, which enable self-driving systems to interpret what they see, reason about complex driving scenarios, and generate driving actions. The platform includes large reasoning models, simulation tools for testing rare and edge-case scenarios, and open datasets for training and validation. NVIDIA says the approach improves transparency, safety, and robustness in autonomous systems, particularly in complex real-world environments, and supports progress toward higher levels of vehicle autonomy: "With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision. Developers can adapt Alpamayo 1 into smaller runtime models for vehicle development, or use it as a foundation for AV development tools such as reasoning-based evaluators and auto-labeling systems. Alpamayo 1 provides open model weights and open-source inferencing scripts. Future models in the family will feature larger parameter counts, more detailed reasoning capabilities, more input and output flexibility, and options for commercial usage."

NEWS: NVIDIA just announced Alpamayo, what CEO Jensen Huang calls the world’s first thinking, reasoning autonomous vehicle AI, launching on U.S. roads later this year, starting with the Mercedes CLA. Jensen: "It's trained end-to-end. Literally from camera in to actuation out; It reasons what action it is about to take, the reason by which is came about that action, and the trajectory." Alpamayo introduces Vision-Language-Action (VLA) models, which enable self-driving systems to interpret what they see, reason about complex driving scenarios, and generate driving actions. The platform includes large reasoning models, simulation tools for testing rare and edge-case scenarios, and open datasets for training and validation. NVIDIA says the approach improves transparency, safety, and robustness in autonomous systems, particularly in complex real-world environments, and supports progress toward higher levels of vehicle autonomy: "With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision. Developers can adapt Alpamayo 1 into smaller runtime models for vehicle development, or use it as a foundation for AV development tools such as reasoning-based evaluators and auto-labeling systems. Alpamayo 1 provides open model weights and open-source inferencing scripts. Future models in the family will feature larger parameter counts, more detailed reasoning capabilities, more input and output flexibility, and options for commercial usage."

Sawyer Merritt

1,604,512 views • 8 months ago