Search papers, labs, and topics across Lattice.
3
0
5
8
CausalDS reveals that LLMs can effectively navigate complex causal reasoning tasks in data science, but struggle with uncertainty and abstention in ambiguous scenarios.
Training on compound reasoning traces yields better generalization than isolated atomic modules, reshaping our understanding of how LLMs can learn to reason.
K2-V2 redefines the open-source LLM landscape, rivaling closed-source giants in reasoning capabilities while providing full transparency into its training data and process.
