Spectral Prior for Reducing Exposure Bias in Diffusion Models

• 类型：arxiv
• 标识：2607.22091
• 链接：https://arxiv.org/abs/2607.22091
• 主分类：multimodal
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：Spectral Alignment is proposed, a lightweight, guidance-based method that calibrates the power spectrum of intermediate predictions to a pre-computed prior and is complementary to Classifier-Free Guidance (CFG).
• 待LLM分类：否
• 标题中文：[标题中文] 降低扩散模型 Exposure Bias 的频谱先验
• TLDR中文：提出 Spectral Alignment，一种基于 guidance 的轻量级方法，将中间预测的功率谱校准到预先计算的先验，并与 Classifier-Free Guidance (CFG) 互补。
• 来源文件：
• /inbox/tom/_candidates/2026-07-27-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-07-28-rag-retrieval-reranking-candidates.json
• /inbox/tom/_candidates/2026-07-28-agent-rag-longcontext-candidates.json
• [S2 enrich]