What Semantics Survive the Connector? Diagnosing VLM-to-DiT Alignment in Video Editing

TL;DR

This paper investigates whether the alignment process between vision-language models and DiT models preserves fine-grained semantics in video editing, revealing it often acts as a semantic bottleneck.

Contribution

It introduces TRACE-Edit, a diagnostic dataset and protocol to systematically evaluate semantic preservation in VLM-to-DiT alignment for video editing.

Findings

Alignment degrades fine-grained structural semantics

VLM-to-DiT alignment is a major semantic bottleneck

Proposed diagnostic tools reveal limitations in current models

Abstract

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.