Tim Miller
tim-miller.bsky.social
Tim Miller
@tim-miller.bsky.social
Associate Professor at the Computational Health Informatics Program of Boston Children's Hospital, and Department of Pediatrics at Harvard Medical School. Research interests include natural language processing for biomedical texts, multi-modal ML.
Awesome turnout for what looks like bad weather! Wish I could've been there to celebrate!
November 16, 2025 at 12:28 AM
Thanks crew! Tried it out today and it's amazing how much it already makes Ballard feel more connected.
November 4, 2025 at 12:28 AM
How do these get done? I see a lot in West Woodlawn but none in, e.g., the brewery district a few blocks away. Can we just ask for them? DIY like yellow curb paint? 😂
October 29, 2025 at 6:48 PM
How can one even be shocked without being surprised? Isn't there significant overlap in those two?
October 7, 2025 at 10:14 PM
Looks like he came in and the scoring onslaught began! If he started we would have scored 20.
October 2, 2025 at 11:07 PM
This says so much more about @petebuttigieg.bsky.social than it does about Harrell. Glad it'll get it out there before he tries to run again.
September 22, 2025 at 10:28 PM
The wealthiest man in the world is a child
September 18, 2025 at 9:53 PM
Both of these papers were team efforts, but Wonjin did great work in leading them, congrats! Camera ready versions are in progress that greatly improve these papers in response to reviews. And lots more great work from Wonjin in the pipeline.
August 29, 2025 at 2:34 AM
The second is a short paper (findings), which describes a method for estimating AUROC for LLM-based classifiers. We convert classification prompts into comparison prompts, treat it like a repeated game, and calculate ELO scores for every instance to get probabilities->ROCs. arxiv.org/abs/2502.15018
Using tournaments to calculate AUROC for zero-shot classification with LLMs
Large language models perform surprisingly well on many zero-shot classification tasks, but are difficult to fairly compare to supervised classifiers due to the lack of a modifiable decision boundary....
arxiv.org
August 29, 2025 at 2:34 AM
The first, a long paper accepted in the main conference, is about our work predicting short-term readmission from psychiatric discharge summaries using LLMs in a summarize-then-classify approach. arxiv.org/abs/2502.10388
Aspect-Oriented Summarization for Psychiatric Short-Term Readmission Prediction
Recent progress in large language models (LLMs) has enabled the automated processing of lengthy documents even without supervised training on a task-specific dataset. Yet, their zero-shot performance ...
arxiv.org
August 29, 2025 at 2:34 AM
Lol how weird/lazy is this dude that he thinks using AI is the best way to obtain a picture of himself wearing a button up shirt?
August 2, 2025 at 4:25 AM
I just want someone to read this to Brian verbatim and wait for a response.
July 19, 2025 at 5:35 AM
It was awesome for people watching, really glad I made time to get to a cwc game since I was out of town during the sounders games.
June 26, 2025 at 11:23 PM
I'm stopping by regularly to try to be one of the first to use the new crossing at 51st!
June 4, 2025 at 5:08 PM
Where do you recommend bike parking? Does it feel safe?
May 26, 2025 at 6:26 PM