Dean Carignan

Harsha Nori, Yin Tat Lee, Sheng Zhang et al. · microsoft-research

Generalist foundation models such as GPT-4 have displayed surprising capabilities in a wide variety of domains and tasks. Yet, there is a prevalent assumption that they cannot match specialist capabilities of fine-tuned models. For example, most explorations to date on medical competency benchmarks have leveraged domain-specific training, as exemplified by efforts on BioGPT and Med-PaLM. We build on a prior study of GPT-4's capabilities on medical challenge benchmarks in the absence of special training. Rather than using simple prompting to highlight the model's out-of-the-box capabilities, we perform a systematic exploration of prompt engineering. We find that prompting innovation can unlock deeper specialist capabilities and show that GPT-4 easily tops prior leading results for medical benchmarks. The prompting methods we explore are general purpose, and make no specific use of domain expertise, removing the need for expert-curated content. Our experimental design carefully controls for overfitting during the prompt engineering process. We introduce Medprompt, based on a composition of several prompting strategies. With Medprompt, GPT-4 achieves state-of-the-art results on all nine of the benchmark datasets in the MultiMedQA suite. The method outperforms leading specialist models such as Med-PaLM 2 by a significant margin with an order of magnitude fewer calls to the model. Steering GPT-4 with Medprompt achieves a 27% reduction in error rate on the MedQA dataset over the best methods to date achieved with specialist models and surpasses a score of 90% for the first time. Beyond medical problems, we show the power of Medprompt to generalize to other domains and provide evidence for the broad applicability of the approach via studies of the strategy on exams in electrical engineering, machine learning, philosophy, accounting, law, nursing, and clinical psychology.

Harsha Nori, Nicholas King, Scott Mayer McKinney et al.

Large language models (LLMs) have demonstrated remarkable capabilities in natural language understanding and generation across various domains, including medicine. We present a comprehensive evaluation of GPT-4, a state-of-the-art LLM, on medical competency examinations and benchmark datasets. GPT-4 is a general-purpose model that is not specialized for medical problems through training or engineered to solve clinical tasks. Our analysis covers two sets of official practice materials for the USMLE, a three-step examination program used to assess clinical competency and grant licensure in the United States. We also evaluate performance on the MultiMedQA suite of benchmark datasets. Beyond measuring model performance, experiments were conducted to investigate the influence of test questions containing both text and images on model performance, probe for memorization of content during training, and study probability calibration, which is of critical importance in high-stakes applications like medicine. Our results show that GPT-4, without any specialized prompt crafting, exceeds the passing score on USMLE by over 20 points and outperforms earlier general-purpose models (GPT-3.5) as well as models specifically fine-tuned on medical knowledge (Med-PaLM, a prompt-tuned version of Flan-PaLM 540B). In addition, GPT-4 is significantly better calibrated than GPT-3.5, demonstrating a much-improved ability to predict the likelihood that its answers are correct. We also explore the behavior of the model qualitatively through a case study that shows the ability of GPT-4 to explain medical reasoning, personalize explanations to students, and interactively craft new counterfactual scenarios around a medical case. Implications of the findings are discussed for potential uses of GPT-4 in medical education, assessment, and clinical practice, with appropriate attention to challenges of accuracy and safety.

Brian Houck, Tim Bozarth, David Liu et al.

Frameworks such as SPACE, DevEx, and DORA established that developer productivity is inherently multidimensional, but left practitioners with a practical question: what should we measure, and how should we use it to improve? This paper introduces Engineering Thrive (EngThrive), a measurement and improvement system developed and deployed across Microsoft's engineering organization. EngThrive organizes productivity around three dimensions - Speed, Ease, and Quality - with Thriving as a guardrail to ensure developer wellbeing improves alongside performance. Within each dimension, outcome-oriented North Star metrics are paired with diagnostic submetrics, combining system telemetry with developer surveys to provide both scale and context. We describe the design principles that guide metric selection, including an approach in which well-chosen metrics align "gaming" behavior with genuine improvement. We also outline the data platform, survey program, and dashboard ecosystem required to operationalize this approach in practice, and present case studies demonstrating how outcome-oriented measurement enables sustained, system-level improvements. Finally, we show that EngThrive functions as a general-purpose evaluation language, applicable not only to developer tools and AI, but to organizational policies, work environments, and other factors that shape how developers experience their work. We offer EngThrive as a concrete model for organizations seeking to move beyond measuring activity toward improving outcomes.

Kyle O'Brien, David Majercak, Xavier Fernandes et al.

Responsible deployment of language models requires mechanisms for refusing unsafe prompts while preserving model performance. While most approaches modify model weights through additional training, we explore an alternative: steering model activations at inference time via amplifying sparse autoencoder (SAE) features that mediate refusal. This work uncovers a fundamental tension between SAE steering-based safety improvements and general model capabilities. While feature steering successfully improves robustness against both single-turn and challenging multi-turn jailbreak attempts, we discover that this comes at a previously underexplored cost -- systematic degradation of performance across multiple benchmark tasks, even on safe inputs with no apparent connection to refusal behavior. This suggests that features mediating refusal may be more deeply entangled with general language model capabilities than previously understood. Our findings reveal important open questions about the nature of safety-relevant features in language models and the feasibility of isolating them for targeted intervention. While SAE-based steering shows promise as a flexible approach to enhancing language model safety, our results highlight the critical need to understand and address the mechanisms behind these capability tradeoffs before such techniques can be practically deployed.

Maxime Peyrard, Sarvjeet Singh Ghotra, Martin Josifoski et al.

Large pretrained language models are critical components of modern NLP pipelines. Yet, they suffer from spurious correlations, poor out-of-domain generalization, and biases. Inspired by recent progress in causal machine learning, in particular the invariant risk minimization (IRM) paradigm, we propose invariant language modeling, a framework for learning invariant representations that generalize better across multiple environments. In particular, we adapt a game-theoretic formulation of IRM (IRM-games) to language models, where the invariance emerges from a specific training schedule in which all the environments compete to optimize their own environment-specific loss by updating subsets of the model in a round-robin fashion. We focus on controlled experiments to precisely demonstrate the ability of our method to (i) remove structured noise, (ii) ignore specific spurious correlations without affecting global performance, and (iii) achieve better out-of-domain generalization. These benefits come with a negligible computational overhead compared to standard training, do not require changing the local loss, and can be applied to any language model. We believe this framework is promising to help mitigate spurious correlations and biases in language models.