解锁扩散语言模型的提示填充能力
Unlocking Prompt Infilling Capability for Diffusion Language Models
AI总结：
本文通过扩展监督微调中的全序列掩码，解锁扩散语言模型的提示填充能力，证明模型填充的提示能匹配或超越手动设计模板，并有效跨模型迁移。
摘要：
掩码扩散语言模型（dLMs）通过双向去噪生成文本，但这一能力仍被限制在提示填充中。这一限制是当前监督微调（SFT）传统仅对响应进行掩码的产物。为解锁这一能力，我们扩展了SFT期间的全序列掩码，其中提示和响应共同被掩码。一旦解锁，模型可根据少量示例填充提示模板中的掩码部分。我们证明，此类模型填充的提示可匹配或超越手动设计的模板，并在模型间有效迁移，且与现有提示优化方法互补。我们的结果表明，训练实践，而非架构限制，是阻碍掩码扩散语言模型生成有效提示的主要瓶颈。
英文摘要：
Masked diffusion language models (dLMs) generate text through bidirectional denoising, yet this capability remains locked for infilling prompts. This limitation is an artifact of the current supervised finetuning (SFT) convention of applying response-only masking. To unlock this capability, we extend full-sequence masking during SFT, where both prompts and responses are masked jointly. Once unlocked, the model infills masked portions of a prompt template conditioned on few-shot examples. We show that such model-infilled prompts match or surpass manually designed templates, transfer effectively across models, and are complementary to existing prompt optimization methods. Our results suggest that training practices, not architectural limitations, are the primary bottleneck preventing masked diffusion language models from infilling effective prompts
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。
