Multimodal 🖼️ > 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 — 22k hours worth of samples from instruction videos 🤯 > Dataset: SciCap captioning on scientific documents benchmark dataset is released along with the challenge!
LLMs 💬 > Microsoft released Phi-4, sota open-source 14B language model 🔥 > Dolphin is back with Dolphin 3.0 Llama 3.1 8B 🐬🐬 > 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 💭 > Dataset: QWQ-LONGCOT-500K is the dataset used to train SmallThinker, generated using QwQ-32B-preview 📕 > Dataset: @cfahlgren1 released React Code Instructions: a dataset of code instruction-code pairs 📕 > Dataset: Qwen team is on the roll, they just released CodeElo, a dataset of code preferences 👩🏻💻
Embeddings 🔖 > @MoritzLaurer released zero-shot version of ModernBERT large 👏 > KaLM is a new family of performant multilingual embedding models with MIT license built using Qwen2-0.5B
Image/Video Generation ⏯️ > NVIDIA released Cosmos, a new family of diffusion/autoregressive World Foundation Models generating worlds from images, videos and texts 🔥 > 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 ⏯️
> 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 💬
the output segmentation tokens are passed to SAM2, to sort of match text (captions or semantic classes) to masks ⤵️
> 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 ⏯️
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 📈 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 🔥 > 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 🔥
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute 🔥
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:
📈 Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
🎄 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.
🧭 Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM
After some heated discussion 🔥, we clarify our intent re. storage limits on the Hub
TL;DR: - public storage is free, and (unless blatant abuse) unlimited. We do ask that you consider upgrading to PRO and/or Enterprise Hub if possible - private storage is paid above a significant free tier (1TB if you have a paid account, 100GB otherwise)
Multimodal 🖼️ > Google shipped a PaliGemma 2, new iteration of PaliGemma with more sizes: 3B, 10B and 28B, with pre-trained and captioning variants 👏 > OpenGVLab released InternVL2, seven new vision LMs in different sizes, with sota checkpoint with MIT license ✨ > Qwen team at Alibaba released the base models of Qwen2VL models with 2B, 7B and 72B ckpts
LLMs 💬 > Meta released a new iteration of Llama 70B, Llama3.2-70B trained further > EuroLLM-9B-Instruct is a new multilingual LLM for European languages with Apache 2.0 license 🔥 > Dataset: CohereForAI released GlobalMMLU, multilingual version of MMLU with 42 languages with Apache 2.0 license > Dataset: QwQ-LongCoT-130K is a new dataset to train reasoning models > Dataset: FineWeb2 just landed with multilinguality update! 🔥 nearly 8TB pretraining data in many languages!
Image/Video Generation 🖼️ > Tencent released HunyuanVideo, a new photorealistic video generation model > OminiControl is a new editing/control framework for image generation models like Flux
Audio 🔊 > Indic-Parler-TTS is a new text2speech model made by community
New InternVL drop with a state-of-the-art 78B vision language model with MIT license 🔥 https://huggingface.co/collections/OpenGVLab/internvl-25-673e1019b66e2218f68d7c1c The release comes with seven new vision LMs based on InternViT 300M/6B and Qwen2.5 (0.5B, 3B, 32B, 72B) and InternLM2 (8B, 7B, 20B) in different sizes 78B model is of InternViT 6B and Qwen2.5-72B Instruct, can accomplish variety of tasks 👏 Try here OpenGVLab/InternVL
small but mighty 🔥 you can fine-tune SmolVLM on an L4 with batch size of 4 and it will only take 16.4 GB VRAM 🫰🏻 also with gradient accumulation simulated batch size is 16 ✨ I made a notebook that includes all the goodies: QLoRA, gradient accumulation, gradient checkpointing with explanations on how they work 💝 https://github.com/huggingface/smollm/blob/main/finetuning/Smol_VLM_FT.ipynb
🖼️ Multimodal > At Hugging Face we released SmolVLM, a performant and efficient smol vision language model 💗 > Show Lab released ShowUI-2B: new vision-language-action model to build GUI/web automation agents 🤖 > Rhymes AI has released the base model of Aria: Aria-Base-64K and Aria-Base-8K with their respective context length > ViDoRe team released ColSmolVLM: A new ColPali-like retrieval model based on SmolVLM > Dataset: Llava-CoT-o1-Instruct: new dataset labelled using Llava-CoT multimodal reasoning model📖 > Dataset: LLaVA-CoT-100k dataset used to train Llava-CoT released by creators of Llava-CoT 📕
💬 LLMs > Qwen team released QwQ-32B-Preview, state-of-the-art open-source reasoning model, broke the internet 🔥 > AliBaba has released Marco-o1, a new open-source reasoning model 💥 > NVIDIA released Hymba 1.5B Base and Instruct, the new state-of-the-art SLMs with hybrid architecture (Mamba + transformer)
⏯️ Image/Video Generation > Qwen2VL-Flux: new image generation model based on Qwen2VL image encoder, T5 and Flux for generation > Lightricks released LTX-Video, a new DiT-based video generation model that can generate 24 FPS videos at 768x512 res ⏯️ > Dataset: Image Preferences is a new image generation preference dataset made with DIBT community effort of Argilla 🏷️
Audio > OuteAI released OuteTTS-0.2-500M new multilingual text-to-speech model based on Qwen-2.5-0.5B trained on 5B audio prompt tokens
INTELLECT-1 is the first collaboratively trained 10 billion parameter language model trained from scratch on 1 trillion tokens of English text and code.
Increasingly, LLMs are becoming very useful for helping scale annotation tasks, i.e. labelling and filtering. When combined with the structured generation, this can be a very scalable way of doing some pre-annotation without requiring a large team of human annotators.
