蛋白质语言模型中的推理时间毒性缓解
Inference-Time Toxicity Mitigation in Protein Language Models
AI总结：
本文提出LDA方法，通过放大logit差异在推理阶段缓解蛋白质语言模型的毒性生成问题，同时保持生成质量。
摘要：
蛋白质语言模型（PLMs）正成为从头设计蛋白质的实用工具，但其双用途潜力引发了安全问题。我们展示，针对特定分类群进行领域适应可能会导致有毒蛋白质的生成，即使毒性不是训练目标。为了解决这个问题，我们采用Logit Diff Amplification（LDA）作为PLMs的推理时间控制机制。LDA通过放大基线模型与毒性微调模型之间的logit差异来修改token概率，无需重新训练。在四个分类群中，LDA一致地将预测毒性率（通过ToxDL2测量）降低到分类群微调基线以下，同时保持生物合理性。我们通过Fréchet ESM距离和预测折叠性（pLDDT）评估质量，发现LDA保持与天然蛋白质相似的分布特征和结构可行性（不同于基于激活的引导方法，后者倾向于降低序列属性）。我们的结果表明，LDA为蛋白质生成器提供了一个实用的安全调节器，在缓解诱发毒性的同时保留了生成质量。
英文摘要：
Protein language models (PLMs) are becoming practical tools for de novo protein design, yet their dual-use potential raises safety concerns. We show that domain adaptation to specific taxonomic groups can elicit toxic protein generation, even when toxicity is not the training objective. To address this, we adapt Logit Diff Amplification (LDA) as an inference-time control mechanism for PLMs. LDA modifies token probabilities by amplifying the logit difference between a baseline model and a toxicity-finetuned model, requiring no retraining. Across four taxonomic groups, LDA consistently reduces predicted toxicity rate (measured via ToxDL2) below the taxon-finetuned baseline while preserving biological plausibility. We evaluate quality using Fréchet ESM Distance and predicted foldability (pLDDT), finding that LDA maintains distributional similarity to natural proteins and structural viability (unlike activation-based steering methods that tend to degrade sequence properties). Our results demonstrate that LDA provides a practical safety knob for protein generators that mitigates elicited toxicity while retaining generative quality.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
