AnalysisAI ModelsOctober 15, 2026

RTAformer targets relational semantic inertia in video scene graphs

Read original source →arxiv.org

The transformer separately models persistent entities and changing relations by injecting calibrated query differences and aligning cross-frame instances. It aims to overcome relational semantic inertia and representational drift, evaluated on Action Genome.

1 source

More stories today

Open the live feed