用於交互式放射學報告草擬的離散擴散語言模型

摘要

擴散語言模型通過對標記畫布進行雙向去噪來生成文本，而非從左到右逐一生成標記，如今已能與自迴歸生成相媲美。然而，醫學基礎模型幾乎仍完全依賴自迴歸方法。我們對混合專家擴散語言模型 DiffusionGemma-26B 進行了調整，並在相同的 LoRA 配方下，於醫學視覺問答數據集上，將其與同等大小的自迴歸同類模型 Gemma-4-26B 進行基準測試，評分由對冗長性具魯棒性的大語言模型評判器執行。擴散模型在所有數據集上均達到或超越自迴歸模型，且微調後的模型（活躍參數 3.8B）與前沿視覺語言模型不相上下；其解碼速度亦快 3.5 至 4.4 倍。除性能持平外，擴散模型還具備自迴歸模型所缺乏的起草能力：任意順序的填充。由於畫布採用雙向去噪，放射科醫師可固定報告片段，並讓模型填充其中的文本，這是擴散模型固有的操作，而自迴歸模型在此方面表現不佳。這恰好符合真實報告的需求——這些報告往往簡潔，或在臨床醫師與機構間存在不一致。

English

Diffusion language models, which generate text by denoising a token canvas bidirectionally instead of emitting tokens left to right, have become competitive with autoregressive (AR) generation. Medical foundation models, however, remain almost entirely autoregressive. We adapt a mixture-of-experts diffusion language model, DiffusionGemma-26B, and benchmark it against its same-size AR sibling Gemma-4-26B under an identical LoRA recipe on medical visual question answering datasets, scored by a verbosity-robust LLM judge. Diffusion matches or exceeds AR on all of them, and the finetuned model (3.8B active) is competitive with frontier vision-language models; its decoding is also 3.5-4.4x faster. Beyond this parity, the diffusion model offers a drafting capability AR lacks: any-order infill. Because the canvas is denoised bidirectionally, a radiologist can fix report fragments and have the model fill the text between them, an operation inherent to diffusion but not to autoregression, which is subpar at it. This suits real reports, which are often terse or inconsistent across clinicians and institutions.