Search papers, labs, and topics across Lattice.
Affiliation:
1
0
2
High reasoning accuracy in VLMs doesn't equate to reliability, as shown by GPT-5.2's 96% hallucination rate despite top performance metrics.
