TRACE prompting induces MLLMs to produce textual allocentric 3D representations from video, yielding consistent gains on spatial QA benchmarks across multiple model backbones.
2s") – pos: Estimated [x, y] of the camera. – facing: Cardinal direction and axis (e.g.,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Unleashing Spatial Reasoning in Multimodal Large Language Models via Textual Representation Guided Reasoning
TRACE prompting induces MLLMs to produce textual allocentric 3D representations from video, yielding consistent gains on spatial QA benchmarks across multiple model backbones.