Search papers, labs, and topics across Lattice.
Affiliation:
5
0
7
2
MLLM-based retrieval can dramatically enhance accuracy in visual question answering by focusing on semantic alignment rather than mere visual similarity.
Traditional trajectory-based training collapses policy diversity, but DART-SD's topology-aware approach ensures agents can explore optimal paths without penalty.
ProMSA achieves superior accuracy in KB-VQA by dynamically selecting retrieval strategies, outperforming traditional fixed pipelines.
Forget fixed pipelines: training an agent to *learn* when and how to search for knowledge dramatically improves performance on knowledge-based visual question answering.
AI can augment peer review by automatically extracting claims, positioning them in the literature, and executing code to verify results, revealing inconsistencies between reported and reproduced findings.
