Context-Matched Distillation: Teacher Causality for Autoregressive Video Distillation

• 类型：arxiv
• 标识：2608.13391
• 链接：https://arxiv.org/abs/2608.13391
• 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：Context-Matched Distillation (CMD) is introduced, a causal DMD framework that aligns teacher supervision with the information available when each target is generated, and naturally extends to frame-wise and chunk-wise generation, long video distillation, and camera-conditioned distillation.
• 待LLM分类：否
• 标题中文：上下文匹配蒸馏：用于自回归视频蒸馏的教师因果性
• TLDR中文：提出 Context-Matched Distillation（CMD），一种因果 DMD 框架，将 teacher 监督信号与每个 target 生成时可用信息对齐，并可自然扩展到逐帧和逐 chunk 生成、长视频蒸馏以及相机条件蒸馏。
• 来源文件：
• /inbox/tom/_candidates/2026-08-15-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-16-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-17-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-17-agent-memory-tool-use-candidates.json
• [S2 enrich]