Search papers, labs, and topics across Lattice.
3
1
5
2
LLMs can learn mathematical reasoning far more efficiently by adaptively simplifying problems to address specific weaknesses, rather than just escalating complexity.
By focusing on activation-space tail eigenvectors, Astra unlocks better fine-tuning with fewer parameters, even beating full fine-tuning in some cases.
Forget simplistic synthetic data: ChartVerse generates complex charts and reliable reasoning data from scratch, enabling an 8B model to outperform its 30B teacher in chart reasoning.
