Hoffman-Yee Research Grants

Status

Closed

Date

Call for proposals will open in Winter 2025

Topics

Healthcare

Link copied to clipboard!

Stanford HAI Awards $2.75M in Hoffman-Yee Grants
Shana Lynch
Aug 18
announcement
This year’s winners propose innovative, bold ideas pushing the boundaries of artificial intelligence.
Stanford HAI Announces Hoffman-Yee Grants Recipients for 2024
Nikki Goth Itoi
Aug 21
announcement
Six interdisciplinary research teams received a total of $3 million to pursue groundbreaking ideas in the field of AI.
Stanford HAI Announces Four Hoffman-Yee Grantees
Shana Lynch
Oct 25
announcement
The second round of funding will sponsor teams that leverage AI to focus on real-world problems in health care, education, and society.
Hoffman-Yee Symposium
conferenceSep 21, 2021
September
21
2021
2023 Hoffman-Yee Symposium
conferenceSep 19, 20239:00 AM - 5:30 PM
September
19
2023
Four Research Teams Awarded New Hoffman-Yee Grant Funding
Nov 13
announcement
This year's research spans foundation models, health care algorithms, social values in social media, and improved chip technology.
DSPy: Compiling Declarative Language Model Calls into State-of-the-Art Pipelines
Omar Khattab, Matei Zaharia, Christopher Potts
Jan 16
Research
The ML community is rapidly exploring techniques for prompting language models (LMs) and for stacking them into pipelines that solve complex tasks. Unfortunately, existing LM pipelines are typically implemented using hard-coded “prompt templates”, i.e. lengthy strings discovered via trial and error. Toward a more systematic approach for developing and optimizing LM pipelines, we introduce DSPy, a programming model that abstracts LM pipelines as text transformation graphs, or imperative computational graphs where LMs are invoked through declarative modules. DSPy modules are parameterized, meaning they can learn how to apply compositions of prompting, finetuning, augmentation, and reasoning techniques. We design a compiler that will optimize any DSPy pipeline to maximize a given metric, by creating and collecting demonstrations. We conduct two case studies, showing that succinct DSPy programs can express and optimize pipelines that reason about math word problems, tackle multi-hop retrieval, answer complex questions, and control agent loops. Within minutes of compiling, DSPy can automatically produce pipelines that outperform out-of-the-box few-shot prompting as well as expert-created demonstrations for GPT-3.5 and Llama2-13b-chat. On top of that, DSPy programs compiled for relatively small LMs like 770M parameter T5 and Llama2-13b-chat are competitive with many approaches that rely on large and proprietary LMs like GPT-3.5 and on expert-written prompt chains. DSPy is available at https://github.com/stanfordnlp/dspy
Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs
Omar Khattab, David Broman, Josh Purtell, Michael J Ryan, Krista Opsahl-Ong, Matei Zaharia, Christopher Potts
Nov 14
Research
Language Model Programs, i.e. sophisticated pipelines of modular language model (LM) calls, are increasingly advancing NLP tasks, but they require crafting prompts that are jointly effective for all modules. We study prompt optimization for LM programs, i.e. how to update these prompts to maximize a downstream metric without access to module-level labels or gradients. To make this tractable, we factorize our problem into optimizing the free-form instructions and few-shot demonstrations of every module and introduce several strategies to craft task-grounded instructions and navigate credit assignment across modules. Our strategies include (i) program- and data-aware techniques for proposing effective instructions, (ii) a stochastic mini-batch evaluation function for learning a surrogate model of our objective, and (iii) a meta-optimization procedure in which we refine how LMs construct proposals over time. Using these insights we develop MIPRO, a novel algorithm for optimizing LM programs. MIPRO outperforms baseline optimizers on five of seven diverse multi-stage LM programs using a best-in-class open-source model (Llama-3-8B), by as high as 13% accuracy. We have released our new optimizers and benchmark in DSPy at [http://dspy.ai](http://dspy.ai).
Equitable Implementation of a Precision Digital Health Program for Glucose Management in Individuals with Newly Diagnosed Type 1 Diabetes
Ananta Addala, Franziska K Bishop, Korey Hood, Ming Yeh Lee, Victoria Y Ding, Priya Prahalad, Dessi P Zaharieva, Johannes Ferstad, Manisha Desai, David Scheinker, David Maahs, Ramesh Johari
Jul 30
Research
Few young people with type 1 diabetes (T1D) meet glucose targets. Continuous glucose monitoring improves glycemia, but access is not equitable. We prospectively assessed the impact of a systematic and equitable digital-health-team-based care program implementing tighter glucose targets (HbA1c < 7%), early technology use (continuous glucose monitoring starts <1 month after diagnosis) and remote patient monitoring on glycemia in young people with newly diagnosed T1D enrolled in the Teamwork, Targets, Technology, and Tight Control (4T Study 1). Primary outcome was HbA1c change from 4 to 12 months after diagnosis; the secondary outcome was achieving the HbA1c targets. The 4T Study 1 cohort (36.8% Hispanic and 35.3% publicly insured) had a mean HbA1c of 6.58%, 64% with HbA1c < 7% and mean time in the range (70-180 mg dl-1) of 68% at 1 year after diagnosis. Clinical implementation of the 4T Study 1 met the prespecified primary outcome and improved glycemia without unexpected serious adverse events. The strategies in the 4T Study 1 can be used to implement systematic and equitable care for individuals with T1D and translate to care for other chronic diseases.
Smart Start—Designing Powerful Clinical Trials Using Pilot Study Data
Priya Prahalad, Dessi P Zaharieva, Johannes Ferstad, Manisha Desai, David Scheinker, David Maahs, Emily Fox, Ramesh Johari
Jan 22
Research
BACKGROUND
Digital health interventions may be optimized before evaluation in a randomized clinical trial. Although many digital health interventions are deployed in pilot studies, the data collected are rarely used to refine the intervention and the subsequent clinical trials.
METHODS
We leverage natural variation in patients eligible for a digital health intervention in a remote patient-monitoring pilot study to design and compare interventions for a subsequent randomized clinical trial.
RESULTS
Our approach leverages patient heterogeneity to identify an intervention with twice the estimated effect size of an unoptimized intervention.
CONCLUSIONS
Optimizing an intervention and clinical trial based on pilot data may improve efficacy and increase the probability of success. (Funded by the National Institutes of Health and others; ClinicalTrials.gov number, NCT04336969.)
Internal Fractures: The Competing Logics of Social Media Platforms
Chenyan Jia, Chunchen Xu, Jeanne Tsai, Jeffrey Hancock, Michael Bernstein, Angèle Christin
Aug 21
Research
Social media platforms are too often understood as monoliths with clear priorities. Instead, we analyze them as complex organizations torn between starkly different justifications of their missions. Focusing on the case of Meta, we inductively analyze the company’s public materials and identify three evaluative logics that shape the platform’s decisions: an engagement logic, a public debate logic, and a wellbeing logic. There are clear trade-offs between these logics, which often result in internal conflicts between teams and departments in charge of these different priorities. We examine recent examples showing how Meta rotates between logics in its decision-making, though the goal of engagement dominates in internal negotiations. We outline how this framework can be applied to other social media platforms such as TikTok, Reddit, and X. We discuss the ramifications of our findings for the study of online harms, exclusion, and extraction.
Measuring receptivity to misinformation at scale on a social media platform
Christopher K Tokita, Jonathan Nagler, Joshua A Tucker, Kevin Aslett, Richard Bonneau, William P Godel, Zeve Sanderson, Nathaniel Persily
Sep 10
Research
Measuring the impact of online misinformation is challenging. Traditional measures, such as user views or shares on social media, are incomplete because not everyone who is exposed to misinformation is equally likely to believe it. To address this issue, we developed a method that combines survey data with observational Twitter data to probabilistically estimate the number of users both exposed to and likely to believe a specific news story. As a proof of concept, we applied this method to 139 viral news articles and find that although false news reaches an audience with diverse political views, users who are both exposed and receptive to believing false news tend to have more extreme ideologies. These receptive users are also more likely to encounter misinformation earlier than those who are unlikely to believe it. This mismatch between overall user exposure and receptive user exposure underscores the limitation of relying solely on exposure or interaction data to measure the impact of misinformation, as well as the challenge of implementing effective interventions. To demonstrate how our approach can address this challenge, we then conducted data-driven simulations of common interventions used by social media platforms. We find that these interventions are only modestly effective at reducing exposure among users likely to believe misinformation, and their effectiveness quickly diminishes unless implemented soon after misinformation’s initial spread. Our paper provides a more precise estimate of misinformation’s impact by focusing on the exposure of users likely to believe it, offering insights for effective mitigation strategies on social media.
ReMix: Optimizing Data Mixtures for Large Scale Imitation Learning
Chethan Anand Bhateja, Joey Hejna, Karl Pertsch, Yichen Jiang, Dorsa Sadigh
Sep 05
Research
Increasingly large robotics datasets are being collected to train larger foundation models in robotics. However, despite the fact that data selection has been of utmost importance to scaling in vision and natural language processing (NLP), little work in robotics has questioned what data such models should actually be trained on. In this work we investigate how to weigh different subsets or "domains'' of robotics datasets during pre-training to maximize worst-case performance across all possible downstream domains using distributionally robust optimization (DRO). Unlike in NLP, we find that these methods are hard to apply out of the box due to varying action spaces and dynamics across robots. Our method, ReMix, employs early stopping and action normalization and discretization to counteract these issues. Through extensive experimentation on both the Bridge and OpenX datasets, we demonstrate that data curation can have an outsized impact on downstream performance. Specifically, domain weights learned by ReMix outperform uniform weights by over 40% on average and human-selected weights by over 20% on datasets used to train the RT-X models.