Search papers, labs, and topics across Lattice.
Affiliation:
2
0
5
Top LLMs can achieve medal-equivalent scores on elite science exams, but they falter on visual grounding and long-horizon consistency.
The organization of post-training reasoning data into a cohesive framework reveals crucial insights that could accelerate advancements in large reasoning models.
