Ramakanth Kavuluru
@BioNLProc
Faculty at UKY. Views my own. Work: #BioNLP, #NLProc, medical informatics, #datascience, machine learning, deep nets, LLMs, AI & fairness, health+socialdata
I don’t want to touch Grok with a ten foot pole. Yes, it does disagree with him on some topics, but I guess at times they hard coded it to spew his opinions. With several decent options out there, hope people don’t rely on this.
I replicated this result, that Grok focuses nearly entirely on finding out what Elon thinks in order to align with that, on a fresh Grok 4 chat with no custom instructions. grok.com/share/c2hhcmQt…
An year after this preprint was out, we finally have a paper based on this work at the BioNLP workshop at #ACL2025. We are doing it online as neither my postdoc nor I can make it to the conference. Proceedings version: aclanthology.org/2025.bionlp-1.…
DPO-like methods are used to align LLM outputs with human preferences. Can they make QA/IE better, a very different use case. We show how this can be done by *automatically* generating preference data from initial model errors to DPO a T5 model. Preprint: arxiv.org/abs/2407.14000
🚨 New release: MegaScience The largest & highest-quality post-training dataset for scientific reasoning is now open-sourced (1.25M QA pairs)! 📈 Trained models outperform official Instruct baselines 🔬 Covers 7+ disciplines with university-level textbook-grade QA 📄 Paper:…
I know “sliding into DMs” has a different connotation but in my case I am just texting about potential job opportunities for my trainees. Grateful to those who reply. For others, even in this case, it doesn’t feel good to be “left on read” 😀🙏
Not kidding, just saw this on a job ad, from a company that is not Indian, in what they are looking for in an employee: "A “jugaad” spirit: Creative, scrappy problem-solving and execution under ambiguity." It's high time jugaad is listed in Merriam-Webster 🙂
Interesting results and easy to follow thread for TLDR.
🚨 The era of infinite internet data is ending, So we ask: 👉 What’s the right generative modelling objective when data—not compute—is the bottleneck? TL;DR: ▶️Compute-constrained? Train Autoregressive models ▶️Data-constrained? Train Diffusion models Get ready for 🤿 1/n
🤯
New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
📣 Excited to share our real-world study of an LLM clinical copilot, a collab between @OpenAI and @PendaHealth. Across 39,849 live patient visits, clinicians with AI had a 16% relative reduction in diagnostic errors and a 13% reduction in treatment errors vs. those without. 🧵
🎉 Our paper, GenerationPrograms, which proposes a modular framework for attributable text generation, has been accepted to @COLM_conf! GenerationPrograms produces a program that executes to text, providing an auditable trace of how the text was generated and major gains on…
Excited to share GenerationPrograms! 🚀 How do we get LLMs to cite their sources? GenerationPrograms is attributable by design, producing a program that executes text w/ a trace of how the text was generated! Gains of up to +39 Attribution F1 and eliminates uncited sentences,…
We've just release 100+ intermediate checkpoints and our training logs from SmolLM3-3B training. We hope this can be useful to the researcher working on mech interpret, training dynamics, RL and other topics :) Training logs: -> Usual training loss (the gap in the loss are due…
📢Job Opportunity Research Associate for Reasoning in LLMs, University of Bath, UK (Deadline 05 August 2025) We are looking to hire a highly motivated researcher to work on analysing reasoning in LLMs For more information, see: harishtayyarmadabushi.com/now-hiring-res…
14 years into being a PI, this is the 1st survey paper from my lab. Let the citations flow :-) This one is on NLP methods for oncology with a focus on cancer regitries. I now realize the hard work that goes into surveys seeing my student Isaac toil away. link.springer.com/article/10.100…
Big news for data science in higher ed! 🚀Colab now offers 1-year Pro subscriptions free of charge for verified US students/faculty, interactive Slideshow Mode for lectures, & an AI toggle per notebook. Enhance teaching & learning in the upcoming academic year! Read all about it…
In "Mind Your Step (by Step): Chain‑of‑Thought can Reduce Performance on Tasks where Thinking Makes Humans Worse", we connect human "overthinking" insights to LLM reasoning, offering a new lens on when thinking‑out‑loud backfires. 📄 Read the full paper: arxiv.org/abs/2410.21333…
One of the better posters I saw today at #icml25 This gets at the root of the problems we were thinking about when we conceived and wrote the CoT paper.
My @NYUDataScience colleague, Carlos Fernandez-Granda, released the 700-page textbook «Probability and Statistics for Data Science», where he condenses 10 years of teaching experience at @NYUniversity. 200 exercises, 102 notebooks, 115 videos! 🥳🥳🥳 ps4ds.net
We’ve upgraded ScholarQA, our agent that helps researchers conduct literature reviews efficiently by providing detailed answers. Now, when ScholarQA cites a source, it won’t just tell you which paper it came from–you’ll see the exact quote, highlighted in the original PDF. 🧵
We created SuperBPE🚀, a *superword* tokenizer that includes tokens spanning multiple words. When pretraining at 8B scale, SuperBPE models consistently outperform the BPE baseline on 30 downstream tasks (+8% MMLU), while also being 27% more efficient at inference time.🧵
There are many algorithms for constructing pre-training data mixtures—which one should we use? Turns out: many of them fall under one framework, have similar issues, and can be improved with a straightforward modification. Introducing Aioli! 🧄 1/9
MedSigLIP: create embeddings for medical images and text - 400M text + 400M vision encoder - Useful for classification, semantic image retrieval, and more -Trained with chest X-rays, CT slices, MRI slices, dermatology images, and more. huggingface.co/google/medsigl…