CausalMix fits a causal model on 512 runs of a 0.5B model to estimate CATE, then extrapolates optimal mixtures for an 800K data pool applied to 7B and 4B models, outperforming RegMix.
The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
Jupiter-N is a post-trained version of Nemotron 3 Super that reports gains on Welsh benchmarks, terminal agent tasks, and instruction following while retaining base capabilities, released openly as a template for sovereign cultural AI adaptation.
citing papers explorer
-
CausalMix: Data Mixture as Causal Inference for Language Model Training
CausalMix fits a causal model on 512 runs of a 0.5B model to estimate CATE, then extrapolates optimal mixtures for an 800K data pool applied to 7B and 4B models, outperforming RegMix.
-
Jupiter-N Technical Report
Jupiter-N is a post-trained version of Nemotron 3 Super that reports gains on Welsh benchmarks, terminal agent tasks, and instruction following while retaining base capabilities, released openly as a template for sovereign cultural AI adaptation.