Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2305.05609

The Case Records of ChatGPT: Language Models and Complex Clinical Questions

Timothy Poterucha·Pierre Elias·Christopher M. Haggerty
View original source ↗Read source document ↗Export metadata ↓

Abstract

Background: Artificial intelligence language models have shown promise in various applications, including assisting with clinical decision-making as demonstrated by strong performance of large language models on medical licensure exams. However, their ability to solve complex, open-ended cases, which may be representative of clinical practice, remains unexplored. Methods: In this study, the accuracy of large language AI models GPT4 and GPT3.5 in diagnosing complex clinical cases was investigated using published Case Records of the Massachusetts General Hospital. A total of 50 cases requiring a diagnosis and diagnostic test published from January 1, 2022 to April 16, 2022 were identified. For each case, models were given a prompt requesting the top three specific diagnoses and associated diagnostic tests, followed by case text, labs, and figure legends. Model outputs were assessed in comparison to the final clinical diagnosis and whether the model-predicted test would result in a correct diagnosis. Results: GPT4 and GPT3.5 accurately provided the correct diagnosis in 26% and 22% of cases in one attempt, and 46% and 42% within three attempts, respectively. GPT4 and GPT3.5 provided a correct essential diagnostic test in 28% and 24% of cases in one attempt, and 44% and 50% within three attempts, respectively. No significant differences were found between the two models, and multiple trials with identical prompts using the GPT3.5 model provided similar results. Conclusions: In summary, these models demonstrate potential usefulness in generating differential diagnoses but remain limited in their ability to provide a single unifying diagnosis in complex, open-ended cases. Future research should focus on evaluating model performance in larger datasets of open-ended clinical challenges and exploring potential human-AI collaboration strategies to enhance clinical decision-making.

Explore related subjects

cs.CLstat.AP

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Timothy Poterucha, Pierre Elias, Christopher M. Haggerty. 2023-05-09. The Case Records of ChatGPT: Language Models and Complex Clinical Questions. https://arxiv.org/abs/2305.05609

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

The Metanym Game: An LLM Benchmark Without Ground Truth That Rises With the Models It Measures

We introduce a benchmark that is fully self-contained, needs no ground truth, and rises with the models it measures. Language models compete at making analogies and subjectively grade one another; nothing enters from outside. The benchmark reproduces GPQA Diamond, a keyed benchmark of expert-written questions, at r = 0.98, audited for a leak and found clean. We hypothesize that both benchmarks measure the same thing in different ways: a language model holds its knowledge as archetypal contexts, relationship patterns valid across many topic domains. GPQA instantiates the required knowledge in one domain; the Metanym Game instantiates one archetype into several domains, generating analogies, no reasoning required. The reasoning feature of an LLM hardly changes the game's ratings, while it lifts GPQA, which requires derivations. In the game, a player writes a context template whose slots, filled with a set of keywords from a topic domain, instantiate a factually true description of that domain; the instantiations are each other's metaphors, and keywords filling the same slot are metanyms, metaphorically synonymous. Correctness is settled sentence by sentence. Ground truth is replaced by the SVD of the factual rating matrix: its left and right singular vectors rate the players as judges and as generators, two ratings from one factorisation, to our knowledge a first for an LLM council of peers. On the subjective criteria, judges are weighted by their rating consistency under a swept calibration anchor. Generating and judging are different skills: on this roster the strongest generators were middling judges. A council of the five best issues the official ratings; its contestable seats keep it current, a candidate steering signal for self-improving AI. The paper is accompanied by a validating package that recomputes every number.

cs.CL↗
arXiv2026

SCoNE: Selective Context-aware Neuron Editing for Robust Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) is highly sensitive to retrieval noise: when retrieved documents mix informative and irrelevant context, LLMs are easily distracted, leading to hallucinations. To overcome this, we propose SCoNE (Selective Context-aware Neuron Editing), a training-free model editing approach that improves retrieval noise robustness by selectively strengthening context-aware FFN neurons that are identified by both high attribution and high cross-input variability. SCoNE requires only a small number of mining samples, no fine-tuning, and no inference-time overhead. Across various knowledge-intensive question-answering benchmarks and two LLM backbones, SCoNE consistently outperforms competitive baseline methods. Our code is available at https://github.com/HYU-ARK-Lab/SCoNE.

cs.CL↗
arXiv2026

COT-TTS: Audio Context-Aware Text-to-Speech with Chain-of-Thought Reasoning

Recently, text-to-speech systems have made significant progress in speech expressiveness and controllability. However, the speaking style of generated speech typically relies on clear user-specified instructions. In natural conversations, speaking style should be naturally inferred from the preceding conversational context. Therefore, we propose COT-TTS, a context-aware, reasoning-based text-to-speech task. Given historical conversation audio, target text, and a reference speech, the system should comprehend the conversational context, infer an explicit intermediate reasoning, and finally synthesize the target speech with the specified timbre. To support this task, we constructed a large-scale bilingual conversational speech dataset comprising 9 million training samples, including a high-quality subset of 1 million samples. We further constructed a source-disjoint benchmark with 800 human-verified samples and established strong task-specific baselines. Additionally, we developed end-to-end autoregressive models with parameter sizes of 0.6B and 1.7B, generating emotion-labeled transcripts, editable speech style inferences, and speech tokens. Experimental results show that the proposed model achieves performance comparable to large-scale baseline systems with significantly fewer parameters. At the same time, the model performs well in terms of duration consistency and emotional consistency, and can generate appropriate emotional, stress, and rhythmic variations based on the conversational context. To facilitate future research, we will publicly release the data construction pipeline, dataset, trained models, and related resources. The demo page and additional resources are available at https://luckybian.github.io/COT-TTS

cs.CL↗