Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
Leading MLLMs still struggle to articulate and localize physical changes across revisited scenes, exposing a fundamental blind spot in multi-image spatial grounding that targeted synthetic data can overcome.
