Hugging Face (Twitter)
RT @reach_vb: Pretty cool to see a MIT licensed 15B model competing w/ DeepSeek R1 - how are the vibes? 👀
RT @reach_vb: Pretty cool to see a MIT licensed 15B model competing w/ DeepSeek R1 - how are the vibes? 👀
Hugging Face (Twitter)
RT @ClementDelangue: 🦾Great📷 milestone for open-source robotics: pi0 & pi0.5 by @physical_int are now on @huggingface, fully ported to PyTorch in @LeRobotHF and validated side-by-side with OpenPI for everyone to experiment with, fine-tune & deploy in their robots!
As described by Physical Intelligence, π₀.₅ is a Vision-Language-Action model which represents a significant evolution from π₀ to address a big challenge in robotics: open-world generalization.
While robots can perform impressive tasks in controlled environments, π₀.₅ is designed to generalize to entirely new environments and situations that were never seen during training.
Generalization must occur at multiple levels:
- Physical Level: Understanding how to pick up a spoon (by the handle) or plate (by the edge), even with unseen objects in cluttered environments
- Semantic Level: Understanding task semantics, where to put clothes and shoes (laundry hamper, not on the bed), and what tools...
Перейти на оригинальный пост
RT @ClementDelangue: 🦾Great📷 milestone for open-source robotics: pi0 & pi0.5 by @physical_int are now on @huggingface, fully ported to PyTorch in @LeRobotHF and validated side-by-side with OpenPI for everyone to experiment with, fine-tune & deploy in their robots!
As described by Physical Intelligence, π₀.₅ is a Vision-Language-Action model which represents a significant evolution from π₀ to address a big challenge in robotics: open-world generalization.
While robots can perform impressive tasks in controlled environments, π₀.₅ is designed to generalize to entirely new environments and situations that were never seen during training.
Generalization must occur at multiple levels:
- Physical Level: Understanding how to pick up a spoon (by the handle) or plate (by the edge), even with unseen objects in cluttered environments
- Semantic Level: Understanding task semantics, where to put clothes and shoes (laundry hamper, not on the bed), and what tools...
Перейти на оригинальный пост
This media is not supported in your browser
VIEW IN TELEGRAM
Hugging Face (Twitter)
RT @calebfahlgren: .@pmarca: "My guess is we are going to live in a world in which most aggregate AI is going to be executed probably on smaller form factors and probably most of that is going to be open source" https://twitter.com/collision/status/1973473479061278737#m
RT @calebfahlgren: .@pmarca: "My guess is we are going to live in a world in which most aggregate AI is going to be executed probably on smaller form factors and probably most of that is going to be open source" https://twitter.com/collision/status/1973473479061278737#m
Hugging Face (Twitter)
RT @MaziyarPanahi: just hit 4k followers on @huggingface! 🤗
couldn’t have done it without the incredible open-source AI community 💜
Grateful for your trust, support, and collaboration.
RT @MaziyarPanahi: just hit 4k followers on @huggingface! 🤗
couldn’t have done it without the incredible open-source AI community 💜
Grateful for your trust, support, and collaboration.
This media is not supported in your browser
VIEW IN TELEGRAM
Hugging Face (Twitter)
RT @xenovacom: IBM just released Granite 4.0, their latest series of small language models! These models excel at agentic workflows (tool calling), document analysis, RAG, and more. 🚀
The "Micro" (3.4B) model can even run 100% locally in your browser on WebGPU, powered by 🤗 Transformers.js!
RT @xenovacom: IBM just released Granite 4.0, their latest series of small language models! These models excel at agentic workflows (tool calling), document analysis, RAG, and more. 🚀
The "Micro" (3.4B) model can even run 100% locally in your browser on WebGPU, powered by 🤗 Transformers.js!
Hugging Face (Twitter)
RT @charliebtan: 🚀 New dataset: ManyPeptidesMD
https://huggingface.co/datasets/transferable-samplers/many-peptides-md
🤯 4.3 ms of MD across 21,700 peptides
Huge thanks to @huggingface for hosting 🤗
With @majdi_has, @leonklein26, Saifuddin Syed, @dom_beaini, @mmbronstein, @AlexanderTong7, @k_neklyudov
Read on 👇
RT @charliebtan: 🚀 New dataset: ManyPeptidesMD
https://huggingface.co/datasets/transferable-samplers/many-peptides-md
🤯 4.3 ms of MD across 21,700 peptides
Huge thanks to @huggingface for hosting 🤗
With @majdi_has, @leonklein26, Saifuddin Syed, @dom_beaini, @mmbronstein, @AlexanderTong7, @k_neklyudov
Read on 👇
huggingface.co
transferable-samplers/many-peptides-md · Datasets at Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Hugging Face (Twitter)
RT @Alibaba_Qwen: 🚀 Qwen3-VL-30B-A3B-Instruct & Thinking are here!
Smaller size, same powerhouse performance 💪—packed with all the capabilities of Qwen3-VL!
🔧 With just 3B active params, it’s rivaling GPT-5-Mini & Claude4-Sonnet — and often beating them across STEM, VQA, OCR, Video, Agent tasks, and more.
And that’s not all: we’re also releasing an FP8 version, plus the FP8 of the massive Qwen3-VL-235B-A22B!
Try it out and make your multimodal AI applications run faster!🧠🖼️
Qwen Chat: https://chat.qwen.ai/?models=qwen3-vl-30b-a3b
Github&Cookbooks： https://github.com/QwenLM/Qwen3-VL/blob/main/cookbooks
API: https://www.alibabacloud.com/help/en/model-studio/models#5540e6e52e1xx
Blog： https://qwen.ai/blog?id=99f0335c4ad9ff6153e517418d48535ab6d8afef&from=research.latest-advancements-list
ModelScope: https://modelscope.cn/collections/Qwen3-VL-5c7a94c8cb144b
HuggingFace: https://huggingface.co/collections/Qwen/qwen3-vl-68d2a7c1b8a8afce4ebd2dbe
RT @Alibaba_Qwen: 🚀 Qwen3-VL-30B-A3B-Instruct & Thinking are here!
Smaller size, same powerhouse performance 💪—packed with all the capabilities of Qwen3-VL!
🔧 With just 3B active params, it’s rivaling GPT-5-Mini & Claude4-Sonnet — and often beating them across STEM, VQA, OCR, Video, Agent tasks, and more.
And that’s not all: we’re also releasing an FP8 version, plus the FP8 of the massive Qwen3-VL-235B-A22B!
Try it out and make your multimodal AI applications run faster!🧠🖼️
Qwen Chat: https://chat.qwen.ai/?models=qwen3-vl-30b-a3b
Github&Cookbooks： https://github.com/QwenLM/Qwen3-VL/blob/main/cookbooks
API: https://www.alibabacloud.com/help/en/model-studio/models#5540e6e52e1xx
Blog： https://qwen.ai/blog?id=99f0335c4ad9ff6153e517418d48535ab6d8afef&from=research.latest-advancements-list
ModelScope: https://modelscope.cn/collections/Qwen3-VL-5c7a94c8cb144b
HuggingFace: https://huggingface.co/collections/Qwen/qwen3-vl-68d2a7c1b8a8afce4ebd2dbe
Hugging Face (Twitter)
RT @johnschulman2: Really happy to see people reproducing the result that LoRA rank=1 closely matches full fine-tuning on many RL fine-tuning problems. Here are a couple nice ones:
https://twitter.com/ben_burtenshaw/status/1974191312229577085 https://twitter.com/zzlccc/status/1973612326747336767#m
RT @johnschulman2: Really happy to see people reproducing the result that LoRA rank=1 closely matches full fine-tuning on many RL fine-tuning problems. Here are a couple nice ones:
https://twitter.com/ben_burtenshaw/status/1974191312229577085 https://twitter.com/zzlccc/status/1973612326747336767#m
Hugging Face (Twitter)
RT @ClementDelangue: Cool reproduction of “Lora without regret” from @thinkymachines by @ben_burtenshaw in TRL
RT @ClementDelangue: Cool reproduction of “Lora without regret” from @thinkymachines by @ben_burtenshaw in TRL
Hugging Face (Twitter)
RT @jietang: Finally, our open-source GLM-4.6 are trending no. 1 on HF. Thanks to all for the support. We are working on the next version and stay tuned!
RT @jietang: Finally, our open-source GLM-4.6 are trending no. 1 on HF. Thanks to all for the support. We are working on the next version and stay tuned!
Hugging Face (Twitter)
RT @arena: 🚨 New Top Open Model Update!
A relative newcomer to the Arena, @zai_org's GLM-4.6 takes the clear, undisputed #1 spot for Top Open Model. 🏆
It also ranks #4 overall, which is not an easy feat! The next top open model, DeepSeek R1 0528, has been the standing champion for months, now trailing nine points behind.
Congrats to the @zai_org team on this achievement! 🙌 https://twitter.com/Zai_org/status/1973034639708344767#m
RT @arena: 🚨 New Top Open Model Update!
A relative newcomer to the Arena, @zai_org's GLM-4.6 takes the clear, undisputed #1 spot for Top Open Model. 🏆
It also ranks #4 overall, which is not an easy feat! The next top open model, DeepSeek R1 0528, has been the standing champion for months, now trailing nine points behind.
Congrats to the @zai_org team on this achievement! 🙌 https://twitter.com/Zai_org/status/1973034639708344767#m
This media is not supported in your browser
VIEW IN TELEGRAM
Hugging Face (Twitter)
RT @Tu7uruu: Just dropped on HF: kani-tts-370m
A lightweight open-source text-to-speech model that sounds great and runs fast!
> 370M parameters — efficient and deployable on consumer GPUs
> NanoCodec + LFM2-350M
> Natural & expressive voice trained with modern neural TTS techniques
> Fast inference: real-time on a single RTX 3060
RT @Tu7uruu: Just dropped on HF: kani-tts-370m
A lightweight open-source text-to-speech model that sounds great and runs fast!
> 370M parameters — efficient and deployable on consumer GPUs
> NanoCodec + LFM2-350M
> Natural & expressive voice trained with modern neural TTS techniques
> Fast inference: real-time on a single RTX 3060
Hugging Face (Twitter)
RT @rohanpaul_ai: A 7B model, tuned for forms and docs, beats giant models at pulling structured data.
Beats GPT-4.1 on 1,000 extraction tasks, trained for $196.
The team generated synthetic training data that preserves memory across chunks of a long file.
That memory lets the model connect names, dates, and values that appear far apart.
They fine-tuned with Low Rank Adaptation, changing only 0.53% of weights.
They then used Group Relative Policy Optimization with a semantic reward and strict JSON checks.
This setup accepts different surface wording if the meaning matches.
On 1,000 held-out tasks it hit 0.573 mean reward and 89% valid JSON, trained for $196, ahead of GPT-4.1 and others.
Result, a small focused model can outperform general models and cost much less.
----
Paper – arxiv. org/abs/2509.22906
Paper Title: "Extract-0: A Specialized Language Model for Document Information Extraction"
RT @rohanpaul_ai: A 7B model, tuned for forms and docs, beats giant models at pulling structured data.
Beats GPT-4.1 on 1,000 extraction tasks, trained for $196.
The team generated synthetic training data that preserves memory across chunks of a long file.
That memory lets the model connect names, dates, and values that appear far apart.
They fine-tuned with Low Rank Adaptation, changing only 0.53% of weights.
They then used Group Relative Policy Optimization with a semantic reward and strict JSON checks.
This setup accepts different surface wording if the meaning matches.
On 1,000 held-out tasks it hit 0.573 mean reward and 89% valid JSON, trained for $196, ahead of GPT-4.1 and others.
Result, a small focused model can outperform general models and cost much less.
----
Paper – arxiv. org/abs/2509.22906
Paper Title: "Extract-0: A Specialized Language Model for Document Information Extraction"
This media is not supported in your browser
VIEW IN TELEGRAM
Hugging Face (Twitter)
RT @ClementDelangue: Nvidia B200s are now available in @huggingface Inference Endpoints!
The world needs more compute 😅😅😅
RT @ClementDelangue: Nvidia B200s are now available in @huggingface Inference Endpoints!
The world needs more compute 😅😅😅
Hugging Face (Twitter)
RT @zsakib_: When I say HuggingFace in my head I say it with a French accent dropping the H 🤗
RT @zsakib_: When I say HuggingFace in my head I say it with a French accent dropping the H 🤗
Hugging Face (Twitter)
RT @m_olbap: Super excited to finally post this interactive resource! We maintain 1M+ Python LOC across 400+ model architectures in 🤗 Transformers. How do we keep it controlled and keep shipping models?
With @LysandreJik, @pcuenq and @yonigoz we wrote down what makes it possible. Dive here!
RT @m_olbap: Super excited to finally post this interactive resource! We maintain 1M+ Python LOC across 400+ model architectures in 🤗 Transformers. How do we keep it controlled and keep shipping models?
With @LysandreJik, @pcuenq and @yonigoz we wrote down what makes it possible. Dive here!
This media is not supported in your browser
VIEW IN TELEGRAM
Hugging Face (Twitter)
RT @Tu7uruu: Just dropped on HF — NeuTTS Air
Next-gen on-device TTS that matches cloud-level quality while staying fully open source.
> Real-time speech synthesis on CPU/GPU
> 3-second voice cloning, no cloud or data upload
> Compact: under 200 MB, runs on mobile and edge devices
> Multilingual and expressive
> Developed by @neuphonicspeech , optimized for speed and fidelity
RT @Tu7uruu: Just dropped on HF — NeuTTS Air
Next-gen on-device TTS that matches cloud-level quality while staying fully open source.
> Real-time speech synthesis on CPU/GPU
> 3-second voice cloning, no cloud or data upload
> Compact: under 200 MB, runs on mobile and edge devices
> Multilingual and expressive
> Developed by @neuphonicspeech , optimized for speed and fidelity
Hugging Face (Twitter)
RT @julien_c: Super happy to announce that University of Zurich (@UZH_en) just joined @huggingface Academia Hub 🇨🇭🎉
Their students and educators get a better access to collaboration and compute features (including ZeroGPU power) on the Hub. 🔥
RT @julien_c: Super happy to announce that University of Zurich (@UZH_en) just joined @huggingface Academia Hub 🇨🇭🎉
Their students and educators get a better access to collaboration and compute features (including ZeroGPU power) on the Hub. 🔥