一种通用方法，让生成模型更少重复，更快出更多样内容。
A Universal Avoidance Method for Diverse Multi-branch Generation
- 通过惩罚已生成内容的相似性，提升多分支多样性。
- 多样性最高提升1.9倍，速度加快4.4倍，计算量仅为1/64。
- 不依赖特定模型结构，适合扩散与Transformer模型使用。
现代生成模型仍缺乏人类水平的创造力，尤其在多分支多样性方面表现不足。以往方法常伴随高昂计算开销或强依赖模型架构。为此，我们提出UAG（Universal Avoidance Generation），一种模型无关且计算高效的生成策略，通过惩罚先前生成输出间的相似性来增强多样性。该方法可有效提升扩散模型与Transformer模型的多分支生成多样性，额外计算开销极小。实验表明，相比当前最优方法，UAG使多样性最高提升1.9倍，运行速度加快4.4倍，仅需1/64的浮点运算量。完整代码见https://anonymous.4open.science/r/2026_ACL_Universal/。
原文摘要 · Abstract (English)
Modern generative models still lack human-level creativity, particularly in multi-branch diversity. Prior approaches to address this problem often incur heavy computation or strong dependency on model architecture. Therefore, we introduce UAG(Universal Avoidance Generation), a model-agnostic and computationally efficient generation strategy that penalizes similarity among previously generated outputs. Thus, UAG can enhance multi-branch diversity across both diffusion and transformer models, with minimal additional computation. In experiments, our method achieves up to 1.9 times higher diversity, runs 4.4 times faster, and requires only 1/64 of the FLOPs compared to state-of-the-art methods. The full code is https://anonymous.4open.science/r/2026_ACL_Universal/.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。