This week, Adaption Labs launched Invent a Dataset. The characteristic generates a structured, training-ready dataset from a description of the conduct you need a mannequin to be taught. You don’t arrive with a seed corpus, a predefined schema, or a labeling information. Is it deployable? Yes, with one caveat. Invent a Dataset is stay now…
Voice brokers fail on precisely the components of a name that matter most: the order quantity, the callback digits, the e-mail deal with the caller has to jot down down. Gradium AI has launched a brand new text-to-speech mannequin and made it the default throughout its API and Studio. The firm experiences an 81.0% human-rated…
Google Research has launched TimesFM-3, a 330 million parameter time collection basis mannequin that forecasts a number of associated collection in a single ahead cross. Every TimesFM checkpoint via 2.5 was univariate: one collection, its personal historical past, nothing else. TimesFM-3 is pretrained natively for multivariate forecasting on greater than 1 trillion time factors, and…
Two frontier open-weight fashions shipped inside a day of one another this week. Z.ai released GLM-5.3-Flash, a 320B-parameter multimodal MoE mannequin with 18B lively parameters. Alibaba’s Qwen team released Qwen3.8-Flash-Next, a 125B mannequin with 6B lively parameters that previews the Qwen4 structure. The two groups designed these programs independently. Yet their configs learn like near-copies…
In this tutorial, we use Anthropic’s claude-protein-binder-design dataset, which incorporates 1,440 AI-designed miniprotein binders examined in opposition to 16 targets. Because the discharge contains each computational predictions and actual wet-lab outcomes from two impartial labs, we will transcend merely finding out the designs. We consider how nicely construction predictors determine profitable binders, whether or not…
In this tutorial, we discover AutoFigure as a sensible toolkit for producing scientific figures straight from textual content descriptions, paper-like content material, and structured methodological explanations. In this tutorial, we arrange the whole AutoFigure setting, repair dependency points reminiscent of Pillow compatibility, and put together the required rendering instruments for SVG and PNG outputs. We…
In this tutorial, we implement an end-to-end supervised fine-tuning pipeline for the XYZ-Aquila-SFT dataset, Hugging Face Transformers, PyTorch, and PEFT. We stream and examine the dataset, parse multi-turn tool-use trajectories, extract structured instrument calls, analyze corpus traits, and protect embedded reasoning and remark patterns. We then convert instrument schemas between message-embedded and structured codecs, render…
In this tutorial, we construct an end-to-end workflow for working with the SupraLabs reasoning corpus. We stream a consultant subset instantly from the Hugging Face Hub, examine its supply distribution, token-length patterns, activity composition, and reasoning-to-answer ratios, and then apply a sequence of high quality filters to take away unsuitable coaching examples. We rework the…
Yesterday, Liquid AI launched LFM2.5-VL-3B. It is a 3.1B-parameter vision-language mannequin constructed for on-device deployment. The mannequin reads digital screens throughout cellular, internet, and desktop. It grounds objects to coordinates, parses paperwork and charts, and calls instruments from textual content or picture enter. Liquid AI experiences a mean of 69.4 throughout 28 imaginative and prescient…
In this tutorial, we implement an end-to-end MiniMax-H3 video technology workflow utilizing ComfyUI as a headless inference backend. We configure the atmosphere round GPU reminiscence, disk capability, mannequin precision, decision, length, sampling technique, and a number of technology modes, whereas dynamically deciding on an applicable weight profile primarily based on the accessible {hardware}. We set…