We are reproducing the full DeepSeek R1 data and training pipeline so everybody can use their recipe. Instead of doing it in secret we can do it together in the open!
๐งช Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
๐ง Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
๐ฅ Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Multimodal ๐ฌ - We have released SmolVLM -- tiniest VLMs that come in 256M and 500M, with it's retrieval models ColSmol for multimodal RAG ๐ - UI-TARS are new models by ByteDance to unlock agentic GUI control ๐คฏ in 2B, 7B and 72B - Alibaba DAMO lab released VideoLlama3, new video LMs that come in 2B and 7B - MiniMaxAI released Minimax-VL-01, where decoder is based on MiniMax-Text-01 456B MoE model with long context - Dataset: Yale released a new benchmark called MMVU - Dataset: CAIS released Humanity's Last Exam (HLE) a new challenging MM benchmark
LLMs ๐ - DeepSeek-R1 & DeepSeek-R1-Zero: gigantic 660B reasoning models by DeepSeek, and six distilled dense models, on par with o1 with MIT license! ๐คฏ - Qwen2.5-Math-PRM: new math models by Qwen in 7B and 72B - NVIDIA released AceMath and AceInstruct, new family of models and their datasets (SFT and reward ones too!)
Audio ๐ฃ๏ธ - Llasa is a new speech synthesis model based on Llama that comes in 1B,3B, and 8B - TangoFlux is a new audio generation model trained from scratch and aligned with CRPO
Image/Video/3D Generation โฏ๏ธ - Flex.1-alpha is a new 8B pre-trained diffusion model by ostris similar to Flux - tencent released Hunyuan3D-2, new 3D asset generation from images
smolagents can see ๐ฅ we just shipped vision support to smolagents ๐ค agentic computers FTW
you can now: ๐ป let the agent get images dynamically (e.g. agentic web browser) ๐ pass images at the init of the agent (e.g. chatting with documents, filling forms automatically etc) with few LoC change! ๐คฏ you can use transformers models locally (like Qwen2VL) OR plug-in your favorite multimodal inference provider (gpt-4o, antrophic & co) ๐ค
With the big hype around AI agents these days, I couldnโt stop thinking about how AI agents could truly enhance real-world activities. What sort of applications could we build with those AI agents: agentic RAG? self-correcting text-to-sql? Nah, boringโฆ
Passionate about outdoors, Iโve always dreamed of a tool that could simplify planning mountain trips while accounting for all potential risks. Thatโs why I built ๐๐น๐ฝ๐ถ๐ป๐ฒ ๐๐ด๐ฒ๐ป๐, a smart assistant designed to help you plan safe and enjoyable itineraries in the French Alps and Pyrenees.
Built using Hugging Face's ๐๐บ๐ผ๐น๐ฎ๐ด๐ฒ๐ป๐๐ library, Alpine Agent combines the power of AI with trusted resources like ๐๐ฌ๐ช๐ต๐ฐ๐ถ๐ณ.๐ง๐ณ (https://skitour.fr/) and METEO FRANCE. Whether itโs suggesting a route with moderate difficulty or analyzing avalanche risks and weather conditions, this agent dynamically integrates data to deliver personalized recommendations.
In my latest blog post, I share how I developed this projectโfrom defining tools and integrating APIs to selecting the best LLMs like ๐๐ธ๐ฆ๐ฏ2.5-๐๐ฐ๐ฅ๐ฆ๐ณ-32๐-๐๐ฏ๐ด๐ต๐ณ๐ถ๐ค๐ต, ๐๐ญ๐ข๐ฎ๐ข-3.3-70๐-๐๐ฏ๐ด๐ต๐ณ๐ถ๐ค๐ต, or ๐๐๐-4.
โท๏ธ Curious how AI can enhance adventure planning?โจTry the app and share your thoughts: florentgbelidji/alpine-agent ๐ Want to build your own agents? Whether for cooking, sports training, or other passions, the possibilities are endless. Check out the blog post to learn more: https://huggingface.co./blog/florentgbelidji/alpine-agent
Many thanks to @m-ric for helping on building this tool with smolagents!
๐ Multimodal - MiniCPM-o 2.6 is a new sota any-to-any model by OpenBMB (vision, speech and text!) - VideoChat-Flash-Qwen2.5-2B is new video multimodal models by OpenGVLab that come in sizes 2B & 7B in resolutions 224 & 448 - ByteDance released larger SA2VA that comes in 26B parameters - Dataset: VRC-Bench is a new diverse benchmark for multimodal LLM reasoning performance
๐ฌ LLMs - MiniMax-Text-01 is a new huge language model (456B passive 45.9B active params) by MiniMaxAI with context length of 4M tokens ๐คฏ - Dataset: Sky-T1-data-17k is a diverse dataset used to train Sky-T1-32B - kyutai released Helium-1-Preview-2B is a new small multilingual LM - Wayfarer-12B is a new LLM able to write D&D ๐ง๐ปโโ๏ธ - ReaderLM-v2 is a new HTML parsing model by Jina AI - Dria released, Dria-Agent-a-3B, new agentic coding model (Pythonic function calling) based on Qwen2.5 Coder - Unsloth released Phi-4, faster and memory efficient Llama 3.3
๐ผ๏ธ Vision - MatchAnything is a new foundation model for matching - FitDit is a high-fidelity VTON model based on DiT architecture
๐ฃ๏ธ Audio - OuteTTS-0.3-1B is a new multilingual text-to-speech model with voice cloning and emotion control capabilities
๐ Retrieval - lightblue released a new reranker based on Qwen2.5 LB-reranker-0.5B-v1.0 that can handle 95+ languages - cde-small-v2 is a new sota small retrieval model by @jxm
FineWeb2 is a massive multilingual dataset for pre-training language models. Like any web-scale dataset, it contains low-quality content. How can we improve it?
Over the past months, an amazing community of 400+ annotators has been labelling content quality (using Argilla) across 23 languages through the FineWeb-C initiative.
Today, I'm happy to share the first classifier trained on this data.
๐ What we've built:
- A lightweight classifier that efficiently removes low-quality content - 90%+ precision demonstrated on Danish & Swedish - Can process the 43M+ documents in Danish FineWeb2 with minimal compute
๐ Why this matters: The approach can be reproduced for any of the 23 languages in FineWeb-C (data-is-better-together/fineweb-c). We can improve training data quality at scale without massive compute resources by starting with community annotations and training small, efficient classifiers.
Multimodal ๐ผ๏ธ > ByteDance released SA2VA: a family of vision LMs that can take image, video, text and visual prompts > moondream2 is out with new capabilities like outputting structured data and gaze detection! > Dataset: Alibaba DAMO lab released multimodal textbook โ 22k hours worth of samples from instruction videos ๐คฏ > Dataset: SciCap captioning on scientific documents benchmark dataset is released along with the challenge!
LLMs ๐ฌ > Microsoft released Phi-4, sota open-source 14B language model ๐ฅ > Dolphin is back with Dolphin 3.0 Llama 3.1 8B ๐ฌ๐ฌ > Prime-RL released Eurus-2-7B-PRIME a new language model trained using PRIME alignment > SmallThinker-3B is a new small reasoning LM based on Owen2.5-3B-Instruct ๐ญ > Dataset: QWQ-LONGCOT-500K is the dataset used to train SmallThinker, generated using QwQ-32B-preview ๐ > Dataset: @cfahlgren1 released React Code Instructions: a dataset of code instruction-code pairs ๐ > Dataset: Qwen team is on the roll, they just released CodeElo, a dataset of code preferences ๐ฉ๐ปโ๐ป
Embeddings ๐ > @MoritzLaurer released zero-shot version of ModernBERT large ๐ > KaLM is a new family of performant multilingual embedding models with MIT license built using Qwen2-0.5B
Image/Video Generation โฏ๏ธ > NVIDIA released Cosmos, a new family of diffusion/autoregressive World Foundation Models generating worlds from images, videos and texts ๐ฅ > Adobe released TransPixar: a new text-to-video model that can generate assets with transparent backgrounds (a first!) > Dataset: fal released cosmos-openvid-1m Cosmos-tokenized OpenVid-1M with samples from OpenVid-1M
Others > Prior Labs released TabPFNv2, the best tabular transformer is out for classification and regression > Metagene-1 is a new RNA language model that can be used for pathogen detection, zero-shot embedding and genome understanding
This week a few more languages have got 1,000 annotations for the educational quality of data from HuggingFaceFW/fineweb-2.
Why should you care?
The quality of pre-training data can have a big impact on the performance of downstream language models trained on that data (HuggingFaceFW/blogpost-fineweb-v1).
Being able to filter by educational quality is on way of improving the quality of the data you use for training an LLM. Very importantly this approach can also reduce the amount of data needed for pertaining.
Why not use an LLM?
LLMs can be used to annotate educational quality for a subset of data. This data can then be used to train a smaller encoder only model to label the full dataset. However, this may not work well for languages outside of english. This is where fineweb-c (community) comes in.
The community is annotating the educational quality of fineweb2 data. Currently 114 languages have some annotations. These annotations will enable a number of things:
- Evaluate whether an LLM can label the educational quality for texts in that language well - Directly be used for training quality classifiers - Help discover other rules and huerisitcs for refining fineweb2 further for different languages.
> The models are capable of tasks involving vision-language understanding and visual referrals (referring segmentation) both for images and videos โฏ๏ธ
> The models come in 1B, 4B and 8B and are based on InternVL2.5 for base architecture and Qwen2, Qwen2.5 and InternLM2 for language model part (depending on the checkpoint)
> The model is very interesting, it has different encoders for different modalities each (visual prompt, text prompt, image and video) then it concatenates these to feed into LLM ๐ฌ
the output segmentation tokens are passed to SAM2, to sort of match text (captions or semantic classes) to masks โคต๏ธ
> Their annotation pipeline is also interesting, they seems to use two open large vision LMs to refine the annotations, and have different levels of descriptions to provide consistency.
I was initially pretty sceptical about Meta's Coconut paper [1] because the largest perf gains were reported on toy linguistic problems. However, these results on machine translation are pretty impressive!
* Iteratively sample CoTs from the model, using a mix of different search strategies. This gives you something like Stream of Search via prompting. * Verify correctness of each CoT using GPT-4o (needed because exact match doesn't work well in medicine where there are lots of aliases) * Use GPT-4o to reformat the concatenated CoTs into a single stream that includes smooth transitions like "hmm, wait" etc that one sees in o1 * Use the resulting data for SFT & RL * Use sparse rewards from GPT-4o to guide RL training. They find RL gives an average ~3 point boost across medical benchmarks and SFT on this data already gives a strong improvement.
Applying this strategy to other domains could be quite promising, provided the training data can be formulated with verifiable problems!
The paper has a lot of experiments (they trained 84 models!) about what makes the video LMs work โฏ๏ธ
Try the demo for best setup here https://huggingface.co./spaces/Apollo-LMMs/Apollo-3B they evaluate sampling strategies, scaling laws for models and datasets, video representation and more! > The authors find out that whatever design decision was applied to small models also scale properly when the model and dataset are scaled ๐ scaling dataset has diminishing returns for smaller models > They evaluate frame sampling strategies, and find that FPS sampling is better than uniform sampling, and they find 8-32 tokens per frame optimal > They also compare image encoders, they try a variation of models from shape optimized SigLIP to DINOv2 they find google/siglip-so400m-patch14-384 to be most powerful ๐ฅ > they also compare freezing different parts of models, training all stages with some frozen parts give the best yield
They eventually release three models, where Apollo-3B outperforms most 7B models and Apollo 7B outperforms 30B models ๐ฅ
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute ๐ฅ
How? By combining step-wise reward models with tree search algorithms :)
We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think"
We're open sourcing the full recipe and sharing a detailed blog post.
In our blog post we cover:
๐ Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
๐ Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets.
๐งญ Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM
