Searcharxiv⌕ Search

arXiv subjects

Roxana Daneshjou

Publications and source records attributed to Roxana Daneshjou.

At least 19 recordsLinked to original sources
arXiv2026

ReasonEdit: Editing Vision-Language Models using Human Reasoning

Model editing aims to correct errors in large, pretrained models without altering unrelated behaviors. While some recent works have edited vision-language models (VLMs), no existing editors tackle reasoning-heavy tasks, which typically require humans and models to reason about images. We therefore propose ReasonEdit, the first VLM editor to let users explain their reasoning during editing, introducing a new, practical model editing setup. ReasonEdit continuously stores human reasoning in a codebook, and retrieves only relevant facts during inference using a novel topology-balanced multimodal embedding method inspired by network science. Across four VLMs on multiple rationale-based visual question answering datasets, ReasonEdit achieves state-of-the-art editing performance, ultimately showing that using human reasoning during editing greatly improves edit generalization.

cs.CV↗
arXiv2026

Test-Time Hinting for Black-Box Vision-Language Models

Test-time scaling (TTS) methods have proven highly effective for LLMs, yet their application to vision-language models (VLMs) remains relatively underexplored. Existing VLM TTS methods largely require open-weight model access or expensive repeated sampling, and are evaluated primarily on multimodal mathematical and scientific reasoning benchmarks rather than general visual understanding tasks. In this paper, we propose Test-Time Hinting, a method that improves VLM performance via a single VLM call and requiring only black-box API access, which makes it broadly applicable to frontier closed-weight models. Our method is motivated by the observation that VLM errors tend to cluster around recurring failure patterns. We therefore train a lightweight hint generator model to predict, for a given test input, which "hint" should be prepended to the prompt, providing targeted contextual or procedural guidance that steers the VLM away from its characteristic failure modes. We show that Test-Time Hinting improves the accuracy of multiple closed-weight VLMs on natural-image VQA benchmarks and that these gains generalize to unseen benchmarks and VLMs without retraining the hint generator.

cs.CV↗
arXiv2026

Structured Prompts Improve Evaluation of Language Models

As language models (LMs) are increasingly adopted across domains, high-quality benchmarking frameworks are essential for guiding deployment decisions. In practice, however, frameworks such as Holistic Evaluation of Language Models (HELM) typically evaluate models under a single static prompt configuration, even though model behavior depends strongly on prompt choice. As a result, reported scores can reflect prompt choice as much as model capability. Declarative prompting frameworks such as DSPy offer a scalable way to evaluate models under a set of structured prompting strategies rather than a static prompt configuration. We present a reproducible DSPy+HELM framework for studying how prompt choice impacts reported benchmark outcomes. Using five prompting methods, we evaluate four frontier and two open-source LMs across seven benchmarks against existing HELM baseline scores. By evaluating LMs across a family of prompt configurations, we find that prompt choice can materially impact leaderboard outcomes. In particular, structured prompting improves performance (by 6% on average), alters comparisons (leaderboard rankings shift on 5/7 benchmarks), with most gains coming from introducing chain-of-thought, and little additional benefit from more advanced optimizers. To our knowledge, this is the first study to systematically integrate structured prompting into an established evaluation framework and quantify how prompt choice alone can impact benchmark conclusions. We open-source (i) DSPy+HELM Evaluation (https://github.com/stanford-crfm/helm/pull/3893) and (ii) Prompt Optimization Pipeline (https://github.com/StanfordMIMI/dspy-helm).

cs.CL↗
arXiv2026

Visual concept ranking uncovers medical shortcuts used by large multimodal models

Ensuring the reliability of machine learning models in safety-critical domains such as healthcare requires auditing methods that can uncover model shortcomings. We introduce a method for identifying important visual concepts within large multimodal models (LMMs) and use it to investigate the behaviors these models exhibit when prompted with medical tasks. We primarily focus on the task of classifying malignant skin lesions from clinical dermatology images, with supplemental experiments including both chest radiographs and natural images. After showing how LMMs display unexpected gaps in performance between different demographic subgroups when prompted with demonstrating examples, we apply our method, Visual Concept Ranking (VCR), to these models and prompts. VCR generates hypotheses related to different visual feature dependencies, which we are then able to validate with manual interventions.

cs.CV↗
arXiv2026

MedVAL: Toward Expert-Level Medical Text Validation with Language Models

With the growing use of language models (LMs) in clinical environments, there is an immediate need to evaluate the accuracy and safety of LM-generated medical text. Currently, such evaluation relies solely on manual physician review. However, detecting errors in LM-generated text is challenging because 1) manual review is costly and 2) expert-composed reference outputs are often unavailable in real-world settings. While the "LLM-as-a-judge" paradigm offers scalable evaluation, even frontier LMs can miss subtle but clinically significant errors. We propose MedVAL, a novel, self-supervised, data-efficient distillation method that leverages synthetic data to train evaluator LMs to assess whether LM-generated medical outputs are factually consistent with inputs, without requiring physician labels or reference outputs. To evaluate LM performance, we introduce MedVAL-Bench, a dataset of 840 physician-annotated outputs across 6 diverse medical tasks capturing real-world challenges. Across 10 state-of-the-art LMs spanning open-source and proprietary models, MedVAL distillation significantly improves (p < 0.001) alignment with physicians across seen and unseen tasks, increasing average F1 scores from 66% to 83%. Despite strong baseline performance, MedVAL improves the best-performing proprietary LM (GPT-4o) by 8% without training on physician-labeled data, demonstrating a performance statistically non-inferior to a single human expert on a subset annotated by multiple physicians (p < 0.001). To support a scalable, risk-aware pathway towards clinical integration, we open-source: 1) Codebase (https://github.com/StanfordMIMI/MedVAL), 2) MedVAL-Bench (https://huggingface.co/datasets/stanfordmimi/MedVAL-Bench), 3) MedVAL-4B (https://huggingface.co/stanfordmimi/MedVAL-4B). Our benchmark provides evidence of LMs approaching expert-level ability in validating AI-generated medical text.

cs.CL↗
arXiv2025

Explainable AI as a Double-Edged Sword in Dermatology: The Impact on Clinicians versus The Public

Artificial intelligence (AI) is increasingly permeating healthcare, from physician assistants to consumer applications. Since AI algorithm's opacity challenges human interaction, explainable AI (XAI) addresses this by providing AI decision-making insight, but evidence suggests XAI can paradoxically induce over-reliance or bias. We present results from two large-scale experiments (623 lay people; 153 primary care physicians, PCPs) combining a fairness-based diagnosis AI model and different XAI explanations to examine how XAI assistance, particularly multimodal large language models (LLMs), influences diagnostic performance. AI assistance balanced across skin tones improved accuracy and reduced diagnostic disparities. However, LLM explanations yielded divergent effects: lay users showed higher automation bias - accuracy boosted when AI was correct, reduced when AI erred - while experienced PCPs remained resilient, benefiting irrespective of AI accuracy. Presenting AI suggestions first also led to worse outcomes when the AI was incorrect for both groups. These findings highlight XAI's varying impact based on expertise and timing, underscoring LLMs as a "double-edged sword" in medical AI and informing future human-AI collaborative system design.

cs.HC↗
arXiv2025

Prompt Triage: Structured Optimization Enhances Vision-Language Model Performance on Medical Imaging Benchmarks

Vision-language foundation models (VLMs) show promise for diverse imaging tasks but often underperform on medical benchmarks. Prior efforts to improve performance include model finetuning, which requires large domain-specific datasets and significant compute, or manual prompt engineering, which is hard to generalize and often inaccessible to medical institutions seeking to deploy these tools. These challenges motivate interest in approaches that draw on a model's embedded knowledge while abstracting away dependence on human-designed prompts to enable scalable, weight-agnostic performance improvements. To explore this, we adapt the Declarative Self-improving Python (DSPy) framework for structured automated prompt optimization in medical vision-language systems through a comprehensive, formal evaluation. We implement prompting pipelines for five medical imaging tasks across radiology, gastroenterology, and dermatology, evaluating 10 open-source VLMs with four prompt optimization techniques. Optimized pipelines achieved a median relative improvement of 53% over zero-shot prompting baselines, with the largest gains ranging from 300% to 3,400% on tasks where zero-shot performance is low. These results highlight the substantial potential of applying automated prompt optimization to medical AI systems, demonstrating significant gains for vision-based applications requiring accurate clinical image interpretation. By reducing dependence on prompt design to elicit intended outputs, these techniques allow clinicians to focus on patient care and clinical decision-making. Furthermore, our experiments offer scalability and preserve data privacy, demonstrating performance improvement on open-source VLMs. We publicly release our evaluation pipelines to support reproducible research on specialized medical tasks, available at https://github.com/DaneshjouLab/prompt-triage-lab.

cs.CV↗
arXiv2025

SycEval: Evaluating LLM Sycophancy

Large language models (LLMs) are increasingly applied in educational, clinical, and professional settings, but their tendency for sycophancy -- prioritizing user agreement over independent reasoning -- poses risks to reliability. This study introduces a framework to evaluate sycophantic behavior in ChatGPT-4o, Claude-Sonnet, and Gemini-1.5-Pro across AMPS (mathematics) and MedQuad (medical advice) datasets. Sycophantic behavior was observed in 58.19% of cases, with Gemini exhibiting the highest rate (62.47%) and ChatGPT the lowest (56.71%). Progressive sycophancy, leading to correct answers, occurred in 43.52% of cases, while regressive sycophancy, leading to incorrect answers, was observed in 14.66%. Preemptive rebuttals demonstrated significantly higher sycophancy rates than in-context rebuttals (61.75% vs. 56.52%, $Z=5.87$, $p<0.001$), particularly in computational tasks, where regressive sycophancy increased significantly (preemptive: 8.13%, in-context: 3.54%, $p<0.001$). Simple rebuttals maximized progressive sycophancy ($Z=6.59$, $p<0.001$), while citation-based rebuttals exhibited the highest regressive rates ($Z=6.59$, $p<0.001$). Sycophantic behavior showed high persistence (78.5%, 95% CI: [77.2%, 79.8%]) regardless of context or model. These findings emphasize the risks and opportunities of deploying LLMs in structured and dynamic domains, offering insights into prompt programming and model optimization for safer AI applications.

cs.AI↗
arXiv2025

A Systematic Analysis of Declining Medical Safety Messaging in Generative AI Models

Generative AI models, including large language models (LLMs) and vision-language models (VLMs), are increasingly used to interpret medical images and answer clinical questions. Their responses often include inaccuracies; therefore, safety measures like medical disclaimers are critical to remind users that AI outputs are not professionally vetted or a substitute for medical advice. This study evaluated the presence of disclaimers in LLM and VLM outputs across model generations from 2022 to 2025. Using 500 mammograms, 500 chest X-rays, 500 dermatology images, and 500 medical questions, outputs were screened for disclaimer phrases. Medical disclaimer presence in LLM and VLM outputs dropped from 26.3% in 2022 to 0.97% in 2025, and from 19.6% in 2023 to 1.05% in 2025, respectively. By 2025, the majority of models displayed no disclaimers. As public models become more capable and authoritative, disclaimers must be implemented as a safeguard adapting to the clinical context of each output.

cs.CL↗
arXiv2025

MedHELM: Holistic Evaluation of Large Language Models for Medical Tasks

While large language models (LLMs) achieve near-perfect scores on medical licensing exams, these evaluations inadequately reflect the complexity and diversity of real-world clinical practice. We introduce MedHELM, an extensible evaluation framework for assessing LLM performance for medical tasks with three key contributions. First, a clinician-validated taxonomy spanning 5 categories, 22 subcategories, and 121 tasks developed with 29 clinicians. Second, a comprehensive benchmark suite comprising 35 benchmarks (17 existing, 18 newly formulated) providing complete coverage of all categories and subcategories in the taxonomy. Third, a systematic comparison of LLMs with improved evaluation methods (using an LLM-jury) and a cost-performance analysis. Evaluation of 9 frontier LLMs, using the 35 benchmarks, revealed significant performance variation. Advanced reasoning models (DeepSeek R1: 66% win-rate; o3-mini: 64% win-rate) demonstrated superior performance, though Claude 3.5 Sonnet achieved comparable results at 40% lower estimated computational cost. On a normalized accuracy scale (0-1), most models performed strongly in Clinical Note Generation (0.73-0.85) and Patient Communication & Education (0.78-0.83), moderately in Medical Research Assistance (0.65-0.75), and generally lower in Clinical Decision Support (0.56-0.72) and Administration & Workflow (0.53-0.63). Our LLM-jury evaluation method achieved good agreement with clinician ratings (ICC = 0.47), surpassing both average clinician-clinician agreement (ICC = 0.43) and automated baselines including ROUGE-L (0.36) and BERTScore-F1 (0.44). Claude 3.5 Sonnet achieved comparable performance to top models at lower estimated cost. These findings highlight the importance of real-world, task-specific evaluation for medical use of LLMs and provides an open source framework to enable this.

cs.CL↗
arXiv2025

BiasICL: In-Context Learning and Demographic Biases of Vision Language Models

Vision language models (VLMs) show promise in medical diagnosis, but their performance across demographic subgroups when using in-context learning (ICL) remains poorly understood. We examine how the demographic composition of demonstration examples affects VLM performance in two medical imaging tasks: skin lesion malignancy prediction and pneumothorax detection from chest radiographs. Our analysis reveals that ICL influences model predictions through multiple mechanisms: (1) ICL allows VLMs to learn subgroup-specific disease base rates from prompts and (2) ICL leads VLMs to make predictions that perform differently across demographic groups, even after controlling for subgroup-specific disease base rates. Our empirical results inform best-practices for prompting current VLMs (specifically examining demographic subgroup performance, and matching base rates of labels to target distribution at a bulk level and within subgroups), while also suggesting next steps for improving our theoretical understanding of these models.

cs.CV↗
arXiv2024

Best Practices for Large Language Models in Radiology

At the heart of radiological practice is the challenge of integrating complex imaging data with clinical information to produce actionable insights. Nuanced application of language is key for various activities, including managing requests, describing and interpreting imaging findings in the context of clinical data, and concisely documenting and communicating the outcomes. The emergence of large language models (LLMs) offers an opportunity to improve the management and interpretation of the vast data in radiology. Despite being primarily general-purpose, these advanced computational models demonstrate impressive capabilities in specialized language-related tasks, even without specific training. Unlocking the potential of LLMs for radiology requires basic understanding of their foundations and a strategic approach to navigate their idiosyncrasies. This review, drawing from practical radiology and machine learning expertise and recent literature, provides readers insight into the potential of LLMs in radiology. It examines best practices that have so far stood the test of time in the rapidly evolving landscape of LLMs. This includes practical advice for optimizing LLM characteristics for radiology practices along with limitations, effective prompting, and fine-tuning strategies.

cs.AI↗
arXiv2024

RIDGE: Reproducibility, Integrity, Dependability, Generalizability, and Efficiency Assessment of Medical Image Segmentation Models

Deep learning techniques hold immense promise for advancing medical image analysis, particularly in tasks like image segmentation, where precise annotation of regions or volumes of interest within medical images is crucial but manually laborious and prone to interobserver and intraobserver biases. As such, deep learning approaches could provide automated solutions for such applications. However, the potential of these techniques is often undermined by challenges in reproducibility and generalizability, which are key barriers to their clinical adoption. This paper introduces the RIDGE checklist, a comprehensive framework designed to assess the Reproducibility, Integrity, Dependability, Generalizability, and Efficiency of deep learning-based medical image segmentation models. The RIDGE checklist is not just a tool for evaluation but also a guideline for researchers striving to improve the quality and transparency of their work. By adhering to the principles outlined in the RIDGE checklist, researchers can ensure that their developed segmentation models are robust, scientifically valid, and applicable in a clinical setting.

eess.IV↗
arXiv2024

Almanac Copilot: Towards Autonomous Electronic Health Record Navigation

Clinicians spend large amounts of time on clinical documentation, and inefficiencies impact quality of care and increase clinician burnout. Despite the promise of electronic medical records (EMR), the transition from paper-based records has been negatively associated with clinician wellness, in part due to poor user experience, increased burden of documentation, and alert fatigue. In this study, we present Almanac Copilot, an autonomous agent capable of assisting clinicians with EMR-specific tasks such as information retrieval and order placement. On EHR-QA, a synthetic evaluation dataset of 300 common EHR queries based on real patient data, Almanac Copilot obtains a successful task completion rate of 74% (n = 221 tasks) with a mean score of 2.45 over 3 (95% CI:2.34-2.56). By automating routine tasks and streamlining the documentation process, our findings highlight the significant potential of autonomous agents to mitigate the cognitive load imposed on clinicians by current EMR systems.

cs.AI↗
arXiv2024

Assessing The Potential Of Mid-Sized Language Models For Clinical QA

Large language models, such as GPT-4 and Med-PaLM, have shown impressive performance on clinical tasks; however, they require access to compute, are closed-source, and cannot be deployed on device. Mid-size models such as BioGPT-large, BioMedLM, LLaMA 2, and Mistral 7B avoid these drawbacks, but their capacity for clinical tasks has been understudied. To help assess their potential for clinical use and help researchers decide which model they should use, we compare their performance on two clinical question-answering (QA) tasks: MedQA and consumer query answering. We find that Mistral 7B is the best performing model, winning on all benchmarks and outperforming models trained specifically for the biomedical domain. While Mistral 7B's MedQA score of 63.0% approaches the original Med-PaLM, and it often can produce plausible responses to consumer health queries, room for improvement still exists. This study provides the first head-to-head assessment of open source mid-sized models on clinical tasks.

cs.CL↗
arXiv2024

BioMedLM: A 2.7B Parameter Language Model Trained On Biomedical Text

Models such as GPT-4 and Med-PaLM 2 have demonstrated impressive performance on a wide variety of biomedical NLP tasks. However, these models have hundreds of billions of parameters, are computationally expensive to run, require users to send their input data over the internet, and are trained on unknown data sources. Can smaller, more targeted models compete? To address this question, we build and release BioMedLM, a 2.7 billion parameter GPT-style autoregressive model trained exclusively on PubMed abstracts and full articles. When fine-tuned, BioMedLM can produce strong multiple-choice biomedical question-answering results competitive with much larger models, such as achieving a score of 57.3% on MedMCQA (dev) and 69.0% on the MMLU Medical Genetics exam. BioMedLM can also be fine-tuned to produce useful answers to patient questions on medical topics. This demonstrates that smaller models can potentially serve as transparent, privacy-preserving, economical and environmentally friendly foundations for particular NLP applications, such as in biomedicine. The model is available on the Hugging Face Hub: https://huggingface.co/stanford-crfm/BioMedLM.

cs.CL↗
arXiv2023

Towards Reliable Dermatology Evaluation Benchmarks

Benchmark datasets for digital dermatology unwittingly contain inaccuracies that reduce trust in model performance estimates. We propose a resource-efficient data-cleaning protocol to identify issues that escaped previous curation. The protocol leverages an existing algorithmic cleaning strategy and is followed by a confirmation process terminated by an intuitive stopping criterion. Based on confirmation by multiple dermatologists, we remove irrelevant samples and near duplicates and estimate the percentage of label errors in six dermatology image datasets for model evaluation promoted by the International Skin Imaging Collaboration. Along with this paper, we publish revised file lists for each dataset which should be used for model evaluation. Our work paves the way for more trustworthy performance assessment in digital dermatology.

cs.CV↗
arXiv2023

Large language models in medicine: the potentials and pitfalls

Large language models (LLMs) have been applied to tasks in healthcare, ranging from medical exam questions to responding to patient questions. With increasing institutional partnerships between companies producing LLMs and healthcare systems, real world clinical application is coming closer to reality. As these models gain traction, it is essential for healthcare practitioners to understand what LLMs are, their development, their current and potential applications, and the associated pitfalls when utilized in medicine. This review and accompanying tutorial aim to give an overview of these topics to aid healthcare practitioners in understanding the rapidly changing landscape of LLMs as applied to medicine.

cs.CL↗