ViRAC uses LLM/VLM reasoning, informed by human self-reported rationales, to generate virtual agent head rotations that more closely match human head movement data than a saliency-based baseline.
Gaze prediction in dynamic 360 immersive videos
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.GR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
ViRAC: A Vision-Reasoning Agent Head Movement Control Framework in Arbitrary Virtual Environments
ViRAC uses LLM/VLM reasoning, informed by human self-reported rationales, to generate virtual agent head rotations that more closely match human head movement data than a saliency-based baseline.