Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
12
Models that appear equally accurate can possess vastly different agentic reasoning capabilities, revealing the hidden complexities of LLM performance.
MLLMs can ace the test, but still fail to *see*鈥攖hey often succeed at complex reasoning with symbols while failing at basic symbol recognition, revealing a reliance on linguistic priors over true visual perception.
