These results highlight how our PLUTO-4 foundation models enhance PathAI’s AI-pathology products across digital diagnostics and translational research. We’re excited for the new capabilities PLUTO-4 will unlock for our partners and the community! 📄 Learn more in our technical

Beyond public benchmarks, PLUTO-4 shows real-world impact — 🩺 ~10 % improvement across multiple PathAI products, with strong gains in dermatopathology specimen classification. These advances bring us closer to robust, generalizable FMs for pathology applications. #Dermatology

Large-scale multi-node training is often communication-bound. We optimized ViT-G training with DDP + GPUDirect RDMA, tuning parameters (bucket_cap_mb, gradient_as_bucket_view) to saturate InfiniBand bandwidth. Result: 3× throughput improvement and near-linear scaling across 4

Self-supervised ViT training with DINOv2 is unstable at scale. Key changes that stabilized PLUTO-4 training: 🔹 Use bfloat16 → prevents overflow and NaNs in projection heads 🔹 Add register tokens → capture high-norm activations 🔹 Use large batches (≥1024) → smoother

Comparison of Patch-Sizes 🧩 4S (FlexiViT) lets us tune patch size per task ps-8 excels at tasks needing fine-grained features (ex: nuclear segmentation) ps-16 suits coarse pattern tasks (Gleason patterns) #FlexiViT #FoundationModels

PLUTO-4 introduces two models: 🧩 4S — a compact, high-throughput model adaptable to different tasks. Trained with FlexiViT and RoPE for configurable patch sizes and multiscale feature extraction. 💪 4G — a frontier-scale model for complex tasks and peak performance

PLUTO-4 is trained on a diverse dataset of 551,164 WSIs from 137,144 cases across 50 institutions. The dataset spans 40+ organs, 60+ diseases, and 100+ stain variants, capturing real-world variation in tissue, staining, and scanning systems.

🚀 Excited to share PLUTO-4, our new state-of-the-art foundation models for pathology! 🔬 We’re seeing SoTA performance across multiple public benchmarks (EVA and HEST) — surpassing other leading pathology foundation models. (1/6) #AI #MachineLearning #Pathology

The opportunity to standardize the way we construct data sets is important...If we try to build a data set for every use case, we can set ourselves up to fail. We don't want to build a large reference data set that doesn't get used - @balasubramaniac from @Path_AI #FriendDx

🔍 Interpretable concepts found using SAE - SAE trained on PLUTO embeddings disentangled polysemantic features. Single dimensions captured distinct concepts: ✅ Cell types (e.g., cancer cells, red blood cells) ✅Geometric features (e.g. edge of tissue) ✅ Artifacts (surgical ink)

🔬 Impact This study shows the promise & potantial of SAEs in explaining foundation model behavior for medical imaging. Interpretable features unlock: - Potential for clinical AI 🏥 - New biological insights 🧪 🔗 Read the full work: https://t.co/2R6c7ck4wH #AI #Pathology

Feature evolution across layers - SAEs trained on PLUTO’s intermediate layers revealed: Early layers → Low-level color/texture features 🎨 Later layers → Pathology-relevant biological features 🔬 (e.g., monosemantic plasma cell dimension).

Monosemantic representations - Single SAE dimensions correlate with counts of single cell types. For example, SAE-1736 represents plasma cell abundance exclusively - The findings generalized to: ✅ Out-of-domain datasets (CPTAC) ✅ Different stains (H&E, IHC) ✅ Various scanners

🧠 Background and why does this matter? - Each dimension in PLUTO embeddings represents multiple characteristics of the image, making interpretability difficult. - We trained SAEs to reveal mono-semantic features—dimensions representing single, interpretable biological concepts.

PathAI #MachineLearning engineers have recently published new #AI findings for mechanistic interpretability of PLUTO, a pathology #foundationmodel. Using sparse autoencoders (SAEs), we uncovered biologically meaningful and interpretable features. 🧵 https://t.co/4T8GWU7y9p

At #SABCS24 today, David Rimm @RimmPathology @YaleMed, @DrKimAllison, Reena Philip of @US_FDA, Andy Beck of @Path_AI & Giuseppe Viale of @IEOufficiale, discuss the importance measuring w/quantitative assays #BreastCancer.

We sat down with VP of Product, Ben Glass, to discuss how #AI-powered #pathology is revolutionizing fibrosis quantification in #oncology and to learn more about the development of PathAI’s latest product: PathExplore™ #Fibrosis Read More: https://t.co/x0DQEKtnHL #PathTwitter

New @CancerReserch findings in @JCO_ASCO examine variability in HRD results across assays, suggesting the need for consistent & aligned strategies in biomarker selection for #clinicaltrials & clinical decision-making.  #jcooa Read the manuscript here: https://t.co/NfzGo5Jhqh

Learn more about PathExplore™ Fibrosis and access the demo today! #PathTwitter https://t.co/hectBHS2BQ

These methods can be time-consuming, labor-intensive, and difficult to scale for large studies. By combining the power of AI with standard pathology workflows, PathExplore™ Fibrosis is democratizing access to these crucial biomarkers directly from H&E whole slide images.