Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Video might be the next intelligence substrate. Strikingly, video models are beginning to exhibit the same emergent reasoning behaviors first observed in LLMs—multi-path search, self-correction, and layer specialization. We demystify video reasoning and show it doesn’t happen frame-by-frame, but along diffusion steps. 🔗 📄 So, what's next? ;)

Zhongang Cai

1,381 subscribers

71,036 Aufrufe • vor 6 Monaten •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

15 Kommentare

Profilbild von Xun Huang

Xun Huangvor 6 Monaten

It is well known that x0 prediction in early timestep represents the weighted average of all possible outcomes, but I feel it’s misleading to call it “reasoning”. How is “reasoning” defined here?

Profilbild von Zhongang Cai

Zhongang Caivor 6 Monaten

Great point — we don’t consider the presence of multiple possible outcomes at the start to be reasoning. Rather, we view reasoning as the subsequent selection process: early steps maintain multiple candidates, and later steps progressively refine and prune them toward a consistent solution. In that sense, it is analogous to latent reasoning in LLMs—except that here the process is more directly observable through the diffusion trajectory.

Profilbild von mrkelly

mrkellyvor 6 Monaten

Video reasoning feels like magic until you see the compute bill.

Profilbild von Zhongang Cai

Zhongang Caivor 6 Monaten

Don’t worry, one year ago I thought video models could never be accelerated to run in real time. Now guess what ;)

Profilbild von 0xmusashi

0xmusashivor 6 Monaten

@kellypeilinchan

Profilbild von Zhongang Cai

Zhongang Caivor 6 Monaten

@kellypeilinchan This is already happening ;)

Profilbild von Thomas

Thomasvor 6 Monaten

Interesting

Profilbild von John 🔮🌳 🌎☀️

John 🔮🌳 🌎☀️vor 6 Monaten

Who knew “compression” could turn into AGI haha

Profilbild von Critter

Crittervor 6 Monaten

A picture is worth a thousand words

Profilbild von Zhongang Cai

Zhongang Caivor 6 Monaten

True, and I wonder if words are necessary at all in many visual reasoning scenarios

Profilbild von Mohamed

Mohamedvor 6 Monaten

@CritterObserves @caizhongang for video QA (even for very long videos), we proved here that they can be done end to end on visuals losslessly (without intermediate captioning at all) + logarithmic compute:

Profilbild von Zhongang Cai

Zhongang Caivor 6 Monaten

@CritterObserves Interesting work!

Profilbild von Yuanzhi

Yuanzhivor 5 Monaten

This seems to be like a natural result of bidirectional attention and diffusion; would like to know the observation on AR video gen model, e.g., diffusion forcing?

Profilbild von Zhongang Cai

Zhongang Caivor 5 Monaten

Great idea!

Profilbild von Wang Ruisi

Wang Ruisivor 6 Monaten

Our GitHub repository is now live for discussions and upcoming tool releases. Follow us for the latest updates! GitHub:

Ähnliche Videos

Demis Hassabis on the next 12 Months: - Full multimodal convergence: Models like Gemini will seamlessly take in and output text, images, audio, and video, with cross-pollination that boosts reasoning + creativity. - Breakthrough visual intelligence: Image models like Nano Banana Pro will produce highly accurate infographics and show near-human visual understanding. - Language + video fusion: Video models integrated with LLMs unlock richer analysis, storytelling, and step-by-step visual reasoning. - World models go mainstream like Genie 3 - Agents become reliable

Demis Hassabis on the next 12 Months: - Full multimodal convergence: Models like Gemini will seamlessly take in and output text, images, audio, and video, with cross-pollination that boosts reasoning + creativity. - Breakthrough visual intelligence: Image models like Nano Banana Pro will produce highly accurate infographics and show near-human visual understanding. - Language + video fusion: Video models integrated with LLMs unlock richer analysis, storytelling, and step-by-step visual reasoning. - World models go mainstream like Genie 3 - Agents become reliable

Chubby♨️

89,236 Aufrufe • vor 9 Monaten

Everyone is competing on realism, resolution, and generation quality. And that makes sense. But once the visuals become good enough, what becomes the next competitive edge? Maybe it’s reasoning. Can the system understand what happened in a scene, what should happen next, and keep those relationships consistent? Rhapsody 1.0 ( Induce AI ) is built around that idea, adding reasoning, story-state tracking, and verification around video generation. The next leap in AI video might not be making prettier pixels. It might be making the process more controllable. Worth watching: Mahesh KMB

Everyone is competing on realism, resolution, and generation quality. And that makes sense. But once the visuals become good enough, what becomes the next competitive edge? Maybe it’s reasoning. Can the system understand what happened in a scene, what should happen next, and keep those relationships consistent? Rhapsody 1.0 ( Induce AI ) is built around that idea, adding reasoning, story-state tracking, and verification around video generation. The next leap in AI video might not be making prettier pixels. It might be making the process more controllable. Worth watching: Mahesh KMB

Liam Turner

14,576 Aufrufe • vor 16 Tagen

AI has transformed how video is created. We think the next wave is about understanding it. Over the past few years, we've seen remarkable advances in video generation, editing, avatars, and creative tooling. An increasingly important problem is teaching machines to search, analyze, reason over, and extract insight from video - across massive libraries and live streams alike. We're calling this video intelligence, and we're actively looking to back founders building here. We're most excited about companies pushing on the core capabilities: - Video-native models - multimodal embeddings, temporal reasoning, and retrieval built specifically for video rather than adapted from image or text - Real-time and large-scale pipelines - infrastructure for processing, indexing, and querying video at the speed and scale enterprises actually need - Agentic and reasoning layers - systems that don't just retrieve clips but answer questions, surface anomalies, and take action on what they see The models and infrastructure to make this real are appearing to be crossing a capability threshold right now. Multimodal foundation models are maturing, storage costs have collapsed, and enterprises are sitting on years of unstructured video with no way to use it. That infrastructure unlocks a wide range of applications including media and sports workflows, security and physical operations, enterprise knowledge management, advertising analytics, robotics, and consumer products, where video has historically been dark data. If you're building in video intelligence at the model layer, the platform layer, or in a vertical application, we'd love to talk!

AI has transformed how video is created. We think the next wave is about understanding it. Over the past few years, we've seen remarkable advances in video generation, editing, avatars, and creative tooling. An increasingly important problem is teaching machines to search, analyze, reason over, and extract insight from video - across massive libraries and live streams alike. We're calling this video intelligence, and we're actively looking to back founders building here. We're most excited about companies pushing on the core capabilities: - Video-native models - multimodal embeddings, temporal reasoning, and retrieval built specifically for video rather than adapted from image or text - Real-time and large-scale pipelines - infrastructure for processing, indexing, and querying video at the speed and scale enterprises actually need - Agentic and reasoning layers - systems that don't just retrieve clips but answer questions, surface anomalies, and take action on what they see The models and infrastructure to make this real are appearing to be crossing a capability threshold right now. Multimodal foundation models are maturing, storage costs have collapsed, and enterprises are sitting on years of unstructured video with no way to use it. That infrastructure unlocks a wide range of applications including media and sports workflows, security and physical operations, enterprise knowledge management, advertising analytics, robotics, and consumer products, where video has historically been dark data. If you're building in video intelligence at the model layer, the platform layer, or in a vertical application, we'd love to talk!

Jason Cui

36,205 Aufrufe • vor 4 Monaten

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

AK

83,681 Aufrufe • vor 3 Jahren

Goodnight, 𝕏..·˚ ༘ ☾ ⋆｡˚ ☄︎ Newscast about xAI, SpaceX, Starlink, Neuralink, 𝕏, and Grok. Created using the xAI Video API. I upscaled the final frame to use as the starting point for the next clip's generation, but it only helped marginally. However, I found a paper called "FreeInit: Bridging Initialization Gap in Video Diffusion Models" that I'm going to try tomorrow to fix the drifting problem. The approach: 1. FFT both images (anchor frame + last frame) 2. Low-pass filter on anchor (structure/color/layout) 3. High-pass filter on last frame (details/edges/motion) 4. Combine and IFFT back

Goodnight, 𝕏..·˚ ༘ ☾ ⋆｡˚ ☄︎ Newscast about xAI, SpaceX, Starlink, Neuralink, 𝕏, and Grok. Created using the xAI Video API. I upscaled the final frame to use as the starting point for the next clip's generation, but it only helped marginally. However, I found a paper called "FreeInit: Bridging Initialization Gap in Video Diffusion Models" that I'm going to try tomorrow to fix the drifting problem. The approach: 1. FFT both images (anchor frame + last frame) 2. Low-pass filter on anchor (structure/color/layout) 3. High-pass filter on last frame (details/edges/motion) 4. Combine and IFFT back

tetsuo

97,901 Aufrufe • vor 7 Monaten

This is how you build the next generation of AI applications. You need an idea, a good model architecture, and a great harness. That's what MUZIM did. It's an app that runs local multimodal and vision models offline to parse and index your images, videos, and documents frame by frame. • Super fast on-device processing • On-device video & image understanding • You can also bring your API keys for heavy reasoning • Your data stays where it is / nothing goes to the cloud You can do a couple of things with this: 1. Vibe searching for whatever you remember in photos or raw video clips. The app will jump straight to the exact frame and timestamp of the video. This is search as it should be. 2. Use local-running agents to perform tasks on your files. I show a couple of examples in my video. You can check it out here: Thanks to the MUZIM by OpenSoul team for showing me the tool and partnering with me on this post.

This is how you build the next generation of AI applications. You need an idea, a good model architecture, and a great harness. That's what MUZIM did. It's an app that runs local multimodal and vision models offline to parse and index your images, videos, and documents frame by frame. • Super fast on-device processing • On-device video & image understanding • You can also bring your API keys for heavy reasoning • Your data stays where it is / nothing goes to the cloud You can do a couple of things with this: 1. Vibe searching for whatever you remember in photos or raw video clips. The app will jump straight to the exact frame and timestamp of the video. This is search as it should be. 2. Use local-running agents to perform tasks on your files. I show a couple of examples in my video. You can check it out here: Thanks to the MUZIM by OpenSoul team for showing me the tool and partnering with me on this post.

Santiago

65,953 Aufrufe • vor 1 Monat

Test-time scaling, reasoning, and generally search-like processes clearly drive significant gains in LLMs. Largely owed to the structure of language. One would think the same could apply to non-linguistic domains, like image generation, but that obviously depends on whether the structure of the domain's representation lends itself to search. 1D ordered tokens (e.g., image FlexTok, video FlexTok) seem like a natural fit since they enable a step-by-step coarse-to-fine generation. We investigated that and found they indeed enable search and scale far better with test-time compute than 2D grids. See the visuals on the webpage. Appearing in ICML Conference 2026. 🔗 📄

Test-time scaling, reasoning, and generally search-like processes clearly drive significant gains in LLMs. Largely owed to the structure of language. One would think the same could apply to non-linguistic domains, like image generation, but that obviously depends on whether the structure of the domain's representation lends itself to search. 1D ordered tokens (e.g., image FlexTok, video FlexTok) seem like a natural fit since they enable a step-by-step coarse-to-fine generation. We investigated that and found they indeed enable search and scale far better with test-time compute than 2D grids. See the visuals on the webpage. Appearing in ICML Conference 2026. 🔗 📄

Amir Zamir

15,686 Aufrufe • vor 4 Monaten

Some say UN is following a similar path to UNI and CAKE… but similar in what aspect? 🤔 We think “same same, but different” might be the best way to put it. Watch the video to see why 👇🏻

Some say UN is following a similar path to UNI and CAKE… but similar in what aspect? 🤔 We think “same same, but different” might be the best way to put it. Watch the video to see why 👇🏻

Unich.com 🆓

172,169 Aufrufe • vor 1 Jahr

Our research team just released Flex-Forcing: a video generation method that lets a single model switch between generation methods at inference time. Right now there are two main approaches to video generation. Bidirectional diffusion models attend to every frame at once, holding structure well at the cost of speed. Autoregressive models generate frame by frame, so they stream fast and scale to long clips, but accumulate error and drift over time. Flex-Forcing trains a single model to do both, letting you choose from the range at inference based on your compute budget.

Our research team just released Flex-Forcing: a video generation method that lets a single model switch between generation methods at inference time. Right now there are two main approaches to video generation. Bidirectional diffusion models attend to every frame at once, holding structure well at the cost of speed. Autoregressive models generate frame by frame, so they stream fast and scale to long clips, but accumulate error and drift over time. Flex-Forcing trains a single model to do both, letting you choose from the range at inference based on your compute budget.

NVIDIA AI

31,412 Aufrufe • vor 2 Monaten

Andrej Karpathy explained the 5 shifts that turned LLMs from chatbots into agentic systems: 00:00 - Memory turns chat into personal AI 6:41- Multimodal AI can read the world 16:58 - Thinking models solve harder tasks 24:51 - Search makes LLMs live 30:58 - Tools turn LLMs into workers This is not another video about “prompt engineering.” It is a 40-minute roadmap for the next AI workflow: memory / vision / reasoning / search / tools. Watch today, then read the article below on how to turn LLMs into self-improving agent loops.

Andrej Karpathy explained the 5 shifts that turned LLMs from chatbots into agentic systems: 00:00 - Memory turns chat into personal AI 6:41- Multimodal AI can read the world 16:58 - Thinking models solve harder tasks 24:51 - Search makes LLMs live 30:58 - Tools turn LLMs into workers This is not another video about “prompt engineering.” It is a 40-minute roadmap for the next AI workflow: memory / vision / reasoning / search / tools. Watch today, then read the article below on how to turn LLMs into self-improving agent loops.

Morlex

37,456 Aufrufe • vor 2 Monaten

Computer vision can see what happened — agentic AI explains why it matters and what to do next. Here’s how teams are upgrading video analytics with vision language models: 🔍 Turn video into searchable intelligence 🧠 Add context and reasoning to system alerts ⚡ Summarize complex scenes and answer questions automatically Read the blog to see 3 real-world ways to bring agentic AI to computer vision →

Computer vision can see what happened — agentic AI explains why it matters and what to do next. Here’s how teams are upgrading video analytics with vision language models: 🔍 Turn video into searchable intelligence 🧠 Add context and reasoning to system alerts ⚡ Summarize complex scenes and answer questions automatically Read the blog to see 3 real-world ways to bring agentic AI to computer vision →

NVIDIA AI

14,273 Aufrufe • vor 8 Monaten

Hiring RL Engineer! Started off as a curious project at Lossfunk to push the boundaries of LLMs in social reasoning - we are now building RL environments, data, and benchmarks to simulate more real-world scenarios. If you want to train SoTA RL models over multi-GPUs (H200s/B200s) to unlock next AI frontier, this is for you.

Hiring RL Engineer! Started off as a curious project at Lossfunk to push the boundaries of LLMs in social reasoning - we are now building RL environments, data, and benchmarks to simulate more real-world scenarios. If you want to train SoTA RL models over multi-GPUs (H200s/B200s) to unlock next AI frontier, this is for you.

Satpal Singh Rathore

46,010 Aufrufe • vor 1 Jahr

SITUATION EXPLAINED: A new generation of AI labs are betting that world models are the path to general intelligence that can interact with the real world. We had General Intuition CEO Pim de Witte on MTS to explain what world models are, and why intuitively world models offer a clearer path towards general intelligence than LLMs. For humans, text is a technology for compressing information about the unfolding, three-dimensional world around us and communicating it to other people. When LLMs reason about the world, they do so by reconstructing the world through text. World models work by modeling the three-dimensional world at the architecture level and then allowing agents to interact with that world. Unlike an LLM, which understands space and time through text, world models perform spatial and temporal reasoning natively. General Intuition hopes to use world models to create embodied AI—machines that can reason and act in the real world. But other companies are also pushing the frontier in this space. Odyssey ML today announced a multi-agent world model, Agora-1. Their launch video shows four participants interacting with a real-time world model simulation of GoldenEye, a first-person shooter for the N64. A playable preview of the GoldenEye simulation is available on their site.

SITUATION EXPLAINED: A new generation of AI labs are betting that world models are the path to general intelligence that can interact with the real world. We had General Intuition CEO Pim de Witte on MTS to explain what world models are, and why intuitively world models offer a clearer path towards general intelligence than LLMs. For humans, text is a technology for compressing information about the unfolding, three-dimensional world around us and communicating it to other people. When LLMs reason about the world, they do so by reconstructing the world through text. World models work by modeling the three-dimensional world at the architecture level and then allowing agents to interact with that world. Unlike an LLM, which understands space and time through text, world models perform spatial and temporal reasoning natively. General Intuition hopes to use world models to create embodied AI—machines that can reason and act in the real world. But other companies are also pushing the frontier in this space. Odyssey ML today announced a multi-agent world model, Agora-1. Their launch video shows four participants interacting with a real-time world model simulation of GoldenEye, a first-person shooter for the N64. A playable preview of the GoldenEye simulation is available on their site.

MTS

11,676 Aufrufe • vor 4 Monaten

V3D Video Diffusion Models are Effective 3D Generators Automatic 3D generation has recently attracted widespread attention. Recent methods have greatly accelerated the generation speed, but usually produce less-detailed objects due to limited model capacity or 3D data. Motivated by recent advancements in video diffusion models, we introduce V3D, which leverages the world simulation capacity of pre-trained video diffusion models to facilitate 3D generation. To fully unleash the potential of video diffusion to perceive the 3D world, we further introduce geometrical consistency prior and extend the video diffusion model to a multi-view consistent 3D generator. Benefiting from this, the state-of-the-art video diffusion model could be fine-tuned to generate 360degree orbit frames surrounding an object given a single image. With our tailored reconstruction pipelines, we can generate high-quality meshes or 3D Gaussians within 3 minutes. Furthermore, our method can be extended to scene-level novel view synthesis, achieving precise control over the camera path with sparse input views. Extensive experiments demonstrate the superior performance of the proposed approach, especially in terms of generation quality and multi-view consistency

V3D Video Diffusion Models are Effective 3D Generators Automatic 3D generation has recently attracted widespread attention. Recent methods have greatly accelerated the generation speed, but usually produce less-detailed objects due to limited model capacity or 3D data. Motivated by recent advancements in video diffusion models, we introduce V3D, which leverages the world simulation capacity of pre-trained video diffusion models to facilitate 3D generation. To fully unleash the potential of video diffusion to perceive the 3D world, we further introduce geometrical consistency prior and extend the video diffusion model to a multi-view consistent 3D generator. Benefiting from this, the state-of-the-art video diffusion model could be fine-tuned to generate 360degree orbit frames surrounding an object given a single image. With our tailored reconstruction pipelines, we can generate high-quality meshes or 3D Gaussians within 3 minutes. Furthermore, our method can be extended to scene-level novel view synthesis, achieving precise control over the camera path with sparse input views. Extensive experiments demonstrate the superior performance of the proposed approach, especially in terms of generation quality and multi-view consistency

AK

31,997 Aufrufe • vor 2 Jahren

Economic experts have been warning us about the next financial crisis since 2018. I gave up waiting for it to happen but this might be a sign that we are getting very close. Check out this video by Matt Baker.

Economic experts have been warning us about the next financial crisis since 2018. I gave up waiting for it to happen but this might be a sign that we are getting very close. Check out this video by Matt Baker.

RyanMatta 🇺🇸 🦅

250,160 Aufrufe • vor 4 Monaten

We raised $11M to make video editors obsolete. Not the humans, the tools. Think about the hours you lose searching for assets, moving clips frame by frame, fixing animations, checking exports, and redoing the same edits over and over. Poolday handles the entire video production process, start to finish, in one prompt. It learns your style. Uses your assets. Makes the video. QAs its own work. 100% on-brand videos. 100M+ video edits made for businesses all over the world. And today, anyone can use it. Tell us what video you'd like the agent to produce. The agent will build it for the first 50 people.

Alexei Chemenda

1,251,358 Aufrufe • vor 3 Tagen

TokenFlow: Consistent Diffusion Features for Consistent Video Editing paper page: The generative AI revolution has recently expanded to videos. Nevertheless, current state-of-the-art video models are still lagging behind image models in terms of visual quality and user control over the generated content. In this work, we present a framework that harnesses the power of a text-to-image diffusion model for the task of text-driven video editing. Specifically, given a source video and a target text-prompt, our method generates a high-quality video that adheres to the target text, while preserving the spatial layout and motion of the input video. Our method is based on a key observation that consistency in the edited video can be obtained by enforcing consistency in the diffusion feature space. We achieve this by explicitly propagating diffusion features based on inter-frame correspondences, readily available in the model. Thus, our framework does not require any training or fine-tuning, and can work in conjunction with any off-the-shelf text-to-image editing method. We demonstrate state-of-the-art editing results on a variety of real-world videos.

TokenFlow: Consistent Diffusion Features for Consistent Video Editing paper page: The generative AI revolution has recently expanded to videos. Nevertheless, current state-of-the-art video models are still lagging behind image models in terms of visual quality and user control over the generated content. In this work, we present a framework that harnesses the power of a text-to-image diffusion model for the task of text-driven video editing. Specifically, given a source video and a target text-prompt, our method generates a high-quality video that adheres to the target text, while preserving the spatial layout and motion of the input video. Our method is based on a key observation that consistency in the edited video can be obtained by enforcing consistency in the diffusion feature space. We achieve this by explicitly propagating diffusion features based on inter-frame correspondences, readily available in the model. Thus, our framework does not require any training or fine-tuning, and can work in conjunction with any off-the-shelf text-to-image editing method. We demonstrate state-of-the-art editing results on a variety of real-world videos.

AK

143,469 Aufrufe • vor 3 Jahren

🔎 Frame-by-frame breakdown of ProPixel Video Analysis and Research's video: Shows not ONE but TWO distinct interdimensional beings: possibly a short Grey and a tall Grey. Time-dilation effect makes the taller figure appear to move faster (Hal Puthoff predicted this). We see UFOs zip by, they see us in slow motion. The Tall Figure appears to be wearing a wrist device and a large belt. Common attire in abduction accounts. Psionic session by Michael Battista. This is next-level. Not one entity — TWO. 👽

🔎 Frame-by-frame breakdown of ProPixel Video Analysis and Research's video: Shows not ONE but TWO distinct interdimensional beings: possibly a short Grey and a tall Grey. Time-dilation effect makes the taller figure appear to move faster (Hal Puthoff predicted this). We see UFOs zip by, they see us in slow motion. The Tall Figure appears to be wearing a wrist device and a large belt. Common attire in abduction accounts. Psionic session by Michael Battista. This is next-level. Not one entity — TWO. 👽

High Strangeness

79,049 Aufrufe • vor 2 Monaten

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

elvis

141,023 Aufrufe • vor 1 Jahr