What Semantics Survive the Connector? Diagnosing VLM-to-DiT Alignment in Video Editing
For researchers building instruction-based video editing systems, this work provides a diagnostic foundation to evaluate and improve multi-modal alignment, overturning a key assumption in the field.
The paper challenges the assumption that connector modules in VLM-to-DiT video editing models preserve fine-grained structural semantics, showing through a controlled diagnostic dataset (TRACE-Edit) that these semantics are severely degraded, identifying alignment as a major bottleneck.
Flow matching based video generative models have been increasingly relying on prepended Vision-Language Models (VLMs) to handle complex, instruction-based video editing. The prevailing assumption underlying this paradigm is that a connector module can seamlessly align the VLM's rich multi-modal reasoning with the original text embedding space of DiTs. However, we hypothesize that this alignment acts as a severe semantic bottleneck, degrading fine-grained structural variables. Verifying this is challenging, as end-to-end evaluations conflate alignment failures with generation errors, and natural datasets lack disentangled annotations. To rigorously investigate this, we propose a controlled data processing pipeline based on video composition that results in TRACE-Edit, a diagnostic dataset focusing on relation-based editing. Leveraging this dataset, we propose a comprehensive diagnostic protocol to analyze two important designs of meta-query and connector in the existing video editing models. Systematic evaluation of four representative model cases reveals that fine-grained structural semantics can be severely degraded during alignment. Our findings overturn the assumption of lossless semantic transfer, identifying the VLM-to-DiT alignment as a major bottleneck and providing a new diagnostic foundation for future multi-modal alignment architectures.