Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
2
0
5
High-performance ML models can be reproduced with minimal information, revealing that they thrive in low-complexity regions and defy traditional overfitting concerns.
Counterintuitively, penalizing prediction error in language model rewards—instead of rewarding it—can drastically improve Best-of-N sampling by preventing reward hacking.