Skip to content
assaynotes
EN日本語中文

#Data and training

2 today

Aug 12

Wednesday

  1. Introducing CARE-X: Towards Clinically Useful Radiology VLMs with Auxiliary Supervision, Reward-Aligned Learning, and Tool-Augmented Measurement

    Microsoft Research released CARE-X, a unified chest X-ray vision-language research model for report generation and structured prediction. It rewards clinical correctness using multitask reinforcement learning (DAPO). Generation and dual-inference modes cover lesion presence and negation, localisation, multilabel classification, catheter and tube malposition detection, and localisation of 29 anatomical regions.

Aug 10

Monday

Aug 6

Thursday

Jul 31

Friday

Jul 30

Thursday

Jul 26

Sunday

Jul 22

Wednesday

Jul 16

Thursday

  1. Newer Models, Same Advantage

    DharmaOCR scored 0.925 on a Portuguese OCR benchmark, ahead of Mistral OCR4's 0.798 and Unlimited-OCR's 0.7587. It specialised through two-stage training: supervised fine-tuning on Portuguese corpora, then DPO to stabilise inference. The author argues that concentrating parameters on one language remains a structural advantage despite emerging architectures.

Jul 9

Thursday

Jul 7

Tuesday

Jul 6

Monday

  1. PRX Part 4: Our Data Strategy

    Hugging Face's fourth PRX instalment details its data strategy: mixing public and internal pretraining datasets, regenerating long image captions with a VLM and converting them for streaming training. It uses Lance for construction and filtering and MDS for streaming. After switching to Qwen3-VL, text latents are computed during training, with measured throughput loss of around 3–4%, or roughly one extra day for 30 days of training.

Jul 1

Wednesday

Jun 30

Tuesday

Jun 17

Wednesday

  1. From pixels to planning: Earth AI for nature restoration

    Google Research released vector data converting Farmscapes 2020 high-resolution raster maps into usable inventories of hedgerows, stone walls and coppices across over 130,000 square kilometres of the UK. The framework fine-tunes an RSF Vision-Transformer pretrained on over 300 million global satellite images with around 247 square kilometres of labelled data. Polsby–Popper compactness distinguishes woodland, clusters of trees and hedgerows, with a threshold below 0.5 for linear features.

Jun 8

Monday

Jun 4

Thursday

  1. The next chapter in flood resilience: Open sourcing Google’s hydrology framework

    Google Research open-sourced its hydrological modelling framework on GitHub under Apache 2.0, enabling national weather and hydrology agencies to integrate AI flood forecasting. The Python package uses PyTorch and an LSTM architecture, can train or fine-tune on Caravan data, and includes interactive tutorial notebooks and videos.

May 27

Wednesday

  1. Introducing physics AI at Mistral: the foundation for engineering acceleration.

    After incorporating Emmi AI, Mistral launched physical AI capabilities for AI-native industrial engineering with partners including ASML, Airbus, Safran and Siemens Energy. The model predicts physical fields directly from geometry and boundary conditions in seconds through one forward pass on a single GPU, versus hours to weeks per design variant in traditional CFD/FEM. It accelerates design iteration while retaining traditional solvers for validation and edge cases.

May 23

Saturday

  1. Emmi joins Mistral to accelerate the AI-native industry

    Mistral AI signed a definitive agreement this week to acquire Physics AI pioneer Emmi AI, strengthening its AI transformation services for industrial companies. Founded in Austria, Emmi AI has over 30 researchers and engineers focusing on large engineering models that replace days of computation with real-time simulation and build digital twins. Its co-founders and team will join Mistral's Science and Applied AI teams in May.

May 20

Wednesday

May 16

Saturday

  1. Finding the molecular switches behind new infectious diseases

    Cambridge professor Clare Bryant uses Google Co-Scientist to study molecular mechanisms causing sepsis when pathogens such as influenza cross species. Generated and ranked hypotheses identified a previously overlooked protein and then specific amino acid sites. Her team is building cell lines carrying these mutations to test the hypotheses, expecting work that normally takes two to three years to finish in six months.

May 2

Saturday

  1. Catalyzing scientific impact through global partnerships and open resources

    Google Research said its open-source tools and datasets have enabled over 250,000 researchers and developers worldwide. Genomics tools DeepVariant, DeepConsensus and DeepPolisher have supported processing exomes and whole genomes from 2.5 million people; MedGemma has over 4.8 million downloads. Open Health Stack is deployed in more than ten countries, reaching over 65 million beneficiaries.

Apr 30

Thursday

Apr 22

Wednesday

Apr 16

Thursday

Apr 1

Wednesday

Mar 25

Wednesday

Mar 18

Wednesday

  1. Introducing Forge

    Mistral AI launched Forge, a system for enterprises to build frontier-class AI models on proprietary knowledge, supporting pre-training, post-training and reinforcement learning. It handles dense and MoE architectures and, when needed, multimodal inputs, with training and governance on companies' own infrastructure.

Mar 17

Tuesday

Mar 13

Friday

  1. Identifying Interactions at Scale for LLMs

    Berkeley AI Research proposed SPEX and ProxySPEX to identify key interactions driving LLM outputs at scale in feature, data and model-component attribution. SPEX turns interaction search into sparse recovery using sparsity and low order; ProxySPEX exploits hierarchy to match SPEX with roughly ten times fewer ablations.

Mar 12

Thursday

Mar 7

Saturday

Jan 10

Saturday

  1. Information-Driven Design of Imaging Systems

    Berkeley AI Research proposed a mutual-information framework to evaluate and optimise imaging from noisy measurements and noise models. A NeurIPS 2025 paper validates decoder performance predictions for colour photography, radio astronomy, lensless imaging and microscopy. Designs match end-to-end state-of-the-art approaches with less memory and compute, without task-specific decoder design.