iiinova — where your digital self keeps evolving

IAMJB @iamjbd.bsky.social · 12/06/2025

This work builds on our recent study on Automated Structured Radiology Report Generation (x.com/IAMJBDEL/st...) which introduces the dataset and evaluation framework.

010

IAMJB @iamjbd.bsky.social · 12/06/2025

Huge thanks to the amazing team at Stanford Center for Artificial Intelligence in Medicine and Imaging (AIMI): Johannes Moll, Louisa Fay, @asfandyar_azhar, @SophieOstmeier, Tim Lueth, Sergios Gatidis, @curtlanglotz

110

IAMJB @iamjbd.bsky.social · 12/06/2025

📄 Paper: arxiv.org/abs/2506.00200 🌐 Project Page: stanford-aimi.github.io/structuring... 🤗 Models & Data: huggingface.co/collections... All models and datasets are fully open-source — we hope this contributes to the broader medical AI community! 🤝

huggingface.co

Structuring with Lightweight Models - a StanfordAIMI Collection

110

IAMJB @iamjbd.bsky.social · 12/06/2025

We benchmark lightweight models (<300M params) against state-of-the-art LLMs (up to 70B params), using human-reviewed test data and clinically grounded evaluation metrics. Our results highlight the strong potential of specialized, efficient models in clinical NLP application.

110

IAMJB @iamjbd.bsky.social · 12/06/2025

💥 Excited to share our latest work: Structuring Radiology Reports: Challenging LLMs with Lightweight Models In this study, we explore how small, task-specific encoder-decoder models can rival (and sometimes outperform) much larger LLMs; all while being faster, cheaper, and easier to deploy. ons.

120

IAMJB @iamjbd.bsky.social · 09/06/2025

Paper, soon to appear at #ACL2025 main: arxiv.org/pdf/2505.24223 Project page, with all resources (datasets, models, ontology) and usage notes: stanford-aimi.github.io/srrg.html All models and datasets are publicly available as open-source: huggingface.co/collections...

huggingface.co

Structured Radiology Reports - a StanfordAIMI Collection

021

IAMJB @iamjbd.bsky.social · 09/06/2025

4) We conduct a reader study to create a radiologist-validated test set for both the automated structured radiology report task, as well as utterances disease labels from our new ontology. Finally, external evaluation is conducted using out-of-institution data by @hopprai.

110

IAMJB @iamjbd.bsky.social · 09/06/2025

3) We fine-tune popular RRG system on this restructured findings and impression, namely: - Chexagent @StanfordAIMI - MAIRA-2 @MSFTResearch - RaDialog @TU_Muenchen - Chexpert-plus @StanfordAIMI As well as a BERT architecture for the disease classification system on our new ontology.

100

IAMJB @iamjbd.bsky.social · 09/06/2025

2) Since each reported observation, whether in the findings or impression sections, is expressed as a single utterance (1.5M unique utterances in total), we use a large language model to label each one according to a new ontology comprising 72 critical chest X-ray (CXR) observations.

100

IAMJB @iamjbd.bsky.social · 09/06/2025

1) We leverage LLM to restructure MIMIC-CXR and Chexpert-plus (180K Findings sections and 400K Impression sections) into reports categorized by organ system, under strict rules.

110

IAMJB @iamjbd.bsky.social · 09/06/2025

💥 We unveil our paper accepted at the #ACL2025 Main Conference: Automated Structured Report Generation Let's revisit automated radiology report generation for CXR. Free-form reports make it hard for AI systems to learn accurate generation, and even harder to evaluate. 🧵👇 @StanfordAIMI @hopprai

173

IAMJB @iamjbd.bsky.social · 14/03/2025

Just noticed our lightweight RRG model has been downloaded over 92,000 times this months on 🤗HuggingFace. This model was included in the CheXpert-Plus release and contains just 67 million parameters: huggingface.co/IAMJB/chexpe... Its also a top ranking model on RexRank (rexrank.ai)

huggingface.co

IAMJB/chexpert-mimic-cxr-impression-baseline · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

081

IAMJB @iamjbd.bsky.social · 10/03/2025

🧵 What if AI could learn from millions of unlabeled radiology images and reports—and then flexibly adapt to new clinical tasks? In a new comprehensive review in @radiology_rsna, colleagues at stanford dive into how foundation models (FMs) are set to revolutionize radiology!

0182

IAMJB @iamjbd.bsky.social · 03/02/2025

"Second, we develop budget forcing to control test-time compute by forcefully terminating the model's thinking process or lengthening it by appending "Wait" multiple times to the model's generation when it tries to end." What a trick...

1141

IAMJB @iamjbd.bsky.social · 25/01/2025

Is this the last benchmark before AGI? Humanity's Last Exam (HLE) 🤯 3,000 expert-level questions across 100+ subjects, created by nearly 1,000 subject matter experts globally.

2112

IAMJB @iamjbd.bsky.social · 24/01/2025

𝟱. Working Memory: Compiles long-term and task memory to create the final prompt for the LLM. Typically, 1–3 = Long-Term Memory; 5 = Short-Term Memory. Thoughts on agent memory?👇

020

IAMJB @iamjbd.bsky.social · 24/01/2025

𝟮. Semantic Memory: External/grounding knowledge or self-knowledge, similar to RAG context. 𝟯. Procedural Memory: System setup details like prompts, tools, and guardrails (stored in Git/registries). 𝟰. Task Memory: Info retrieved from long-term storage for immediate tasks.

130

IAMJB @iamjbd.bsky.social · 24/01/2025

𝗔 𝗦𝗶𝗺𝗽𝗹𝗲 𝗚𝘂𝗶𝗱𝗲 𝘁𝗼 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁 𝗠𝗲𝗺𝗼𝗿𝘆 🌟 An agent's memory helps it plan and react by leveraging past interactions or external data via prompt context. Here’s a breakdown: 𝟭. Episodic Memory: Logs past actions/interactions (e.g., stored in a vector database for semantic search).

1101

IAMJB @iamjbd.bsky.social · 19/01/2025

🧩 The future of creativity is elemental. ✨ Kling AI just announced Elements 🌎 First, world building: Craft your characters, environments, props. Plan your motion and VFX. 🎛️ Then, remixing: Bring it all together into a cohesive story.

062

IAMJB @iamjbd.bsky.social · 16/01/2025

Amazing. Agent Roles: ⛳ PhD Agent: Conducts literature reviews, interprets results, writes reports. ⛳ Postdoc Agent: Plans research, designs experiments. ⛳ ML Engineer Agent: Prepares data, writes, optimizes code. ⛳ Professor Agent: Oversees, refines reports.

2102

IAMJB @iamjbd.bsky.social · 15/01/2025

MiniMax-01 is Now Open-Source: Scaling Lightning Attention for the AI Agent Era >> Hybrid linear-softmax attention working very well at large scale and long-context filecdn.minimax.chat/_Arxiv_MiniM...

070

IAMJB @iamjbd.bsky.social · 15/01/2025

first look into what the Qwen team used to develop QwQ arxiv.org/pdf/2501.07301

020

IAMJB @iamjbd.bsky.social · 14/01/2025

Neat: Representing Long Volumetric Video with Temporal Gaussian Hierarchy Contrib: Temporal Gaussian Hierarchy representation for long volumetric video.

070

IAMJB @iamjbd.bsky.social · 13/01/2025

Nice visualization of RAG vs. Agentic RAG

262

IAMJB @iamjbd.bsky.social · 12/01/2025

Neat. Converts images, PDFs, and Office documents to Markdown or JSON using OCR and LLM models, with features for caching, distributed processing, and PII removal

1231

IAMJB @iamjbd.bsky.social · 10/01/2025

volume rendering made easy and free 😍

0110

IAMJB @iamjbd.bsky.social · 08/01/2025

How do you even coordinate this?

0150

IAMJB @iamjbd.bsky.social · 05/01/2025

🚀 PRIME + Eurus-2 beat Qwen2.5-Math-Instruct with 1/10 the data! ✨ Implicit PRM (no labels) 🔄 Online updates, zero overhead 🎯 Token-level rewards + RLOO Scaling up with 3x more data!

070

IAMJB @iamjbd.bsky.social · 04/01/2025

Super interesting read: www.thelancet.com/journals/la...

050

IAMJB @iamjbd.bsky.social · 04/01/2025

2 OLMo 2 Furious captures every lesson learned since OLMo 1, featuring in-depth explorations of: • Stable pretraining • LR annealing, data curricula, and soups • Tulu post-training • Compute infrastructure

031

IAMJB @iamjbd.bsky.social · 31/12/2024

Top 25 Open Source AI models on Hugging Face in 2025

290

IAMJB @iamjbd.bsky.social · 16/12/2024

💥 Gemini 2.0 is on paper-central. Talk with any paper from the 🤗 Hugging Face paper page. Example with GenEx 👇

090

IAMJB @iamjbd.bsky.social · 13/12/2024

EasyRef is on 🤗 Hugging Face After DiffSensei yesterday, @ylecun is once again being style-transferred! (⚡Llama 3.3) Chat with the paper: huggingface.co/spaces/hugg... 🤗 Model: huggingface.co/zongzhuofan... 🤗 Demo: huggingface.co/spaces/zong... 🤗 Paper: huggingface.co/papers/2412...

120

IAMJB @iamjbd.bsky.social · 13/12/2024

SuperCharged Euclid is on 🤗 Hugging Face Also, this is the best paper heading I’ve seen in quite some time. The 'en tête' looks fantastic. (⚡Llama 3.3) Chat with the paper: huggingface.co/spaces/hugg... 🤗 Model: huggingface.co/euclid-mult... 🤗 Dataset: huggingface.co/datasets/eu...

092

IAMJB @iamjbd.bsky.social · 13/12/2024

Phi-4 technical report is out. What are the main contributions, how does it compare to SOTA? Find out 👇 🤗 Chat with the paper (⚡Llama 3.3) huggingface.co/spaces/hugg...

010

IAMJB @iamjbd.bsky.social · 13/12/2024

InternLM-XComposer-2.5-OmniLive is on 🤗 Hugging Face (⚡Llama 3.3) Chat with paper: huggingface.co/spaces/hugg... 🤗 Model: huggingface.co/internlm/in... 🤗 Paper: huggingface.co/papers/2412... 🐍 GitHub: github.com/InternLM/In...

030

IAMJB @iamjbd.bsky.social · 12/12/2024

Started training my new LLM today and everything is going well.

270

IAMJB @iamjbd.bsky.social · 12/12/2024

Gemini 2.0 Flash outperforms 1.5 Pro on key benchmarks at 2X speed Now available in anychat try it out: 🤗 huggingface.co/spaces/akha...

huggingface.co

Anychat - a Hugging Face Space by akhaliq

061

IAMJB @iamjbd.bsky.social · 11/12/2024

⚔️ CodeArena is on 🤗 Hugging Face 🤗 Dataset: huggingface.co/datasets/CS... 💡 An excellent way to get an overview in a glimpse: ask paper-central to summarize the scores in a markdown table (⚡by Llama 3.3!) Try it out for any paper: huggingface.co/spaces/hugg...

010

IAMJB @iamjbd.bsky.social · 10/12/2024

Divot from Tencent - Diffusion-Powered Video Tokenizer 💡 An excellent way to get model performance in a glimpse: ask paper-central to summarize the scores in a markdown table (⚡by Llama 3.3!) Try it out for any paper: huggingface.co/spaces/hugg...

000

IAMJB @iamjbd.bsky.social · 10/12/2024

💡 An excellent way to compare a new LLM performance in a glimpse: ask paper-central to summarize the scores in a markdown table. Here is an excellent example for EXAONE 3.5. Try it out: huggingface.co/spaces/hugg...

020

IAMJB @iamjbd.bsky.social · 09/12/2024

New optimizer APOLLO 🤗 💬 How does it compare to Adam? What are the findings? Chat with the paper and find out: 🤗 Chat: huggingface.co/spaces/hugg... 🤗 Paper-page: huggingface.co/papers/2412... 🐍 GitHub: github.com/zhuhanqing/... 📄 Project Page: zhuhanqing.github.io/APOLLO/

040

IAMJB @iamjbd.bsky.social · 09/12/2024

EXAONE 3.5 is on Hugging Face 🤗 💬 How does it compare to SOTA? What are the contributions? Chat with the paper and find out: 🤗 Chat: huggingface.co/spaces/hugg... 🤗 Paper-page: huggingface.co/papers/2412... 🤗 Models: huggingface.co/LGAI-EXAONE... 🤗 Space: huggingface.co/spaces/LGAI...

030

IAMJB @iamjbd.bsky.social · 09/12/2024

MAmmoTH-VL is on Hugging Face 🤗 💬 How does it compare to SOTA? What are the contributions? Chat with the paper and find out: 🤗 Chat: huggingface.co/spaces/hugg... 🤗 Paper-page: huggingface.co/papers/2412... 🤗 Models: huggingface.co/MAmmoTH-VL/... 🤗 Dataset: huggingface.co/datasets/MA...

042