Search papers, labs, and topics across Lattice.
Univ. Lille, Inria, Centrale Lille
2
0
4
Reasoning Core achieves unprecedented performance in completion-supervised reasoning tasks, outpacing existing procedural datasets and revealing critical design insights for effective training.
Forget finetuning, pre-train your LLM on procedurally generated, solver-verified symbolic reasoning data and watch its reasoning skills soar without sacrificing language modeling quality.
