Search papers, labs, and topics across Lattice.
PSL Research University / Inria France.
3
0
5
7
Current large language models are overconfident, but a new calibration method for eigenvalues could significantly enhance their reliability in real-world applications.
DecompRL enables LLMs to solve complex problems by breaking them down into manageable sub-tasks, achieving a 50x reduction in GPU costs while enhancing solution diversity.
Stop relying on small-scale, inconsistent evaluations: CalArena offers a large-scale, standardized benchmark revealing that smooth calibration functions and dedicated multiclass methods are essential for effective post-hoc calibration.
