Domain Mixture Design via Log-Likelihood Differences for Aligning Language Models with a Target Model

source_context

{
  "surface": "signal_canvas",
  "mode": "paper",
  "query": "Domain Mixture Design via Log-Likelihood Differences for Aligning Language Models with a Target Model",
  "normalized_query": "2603.16622",
  "route": "/signal-canvas/domain-mixture-design-via-log-likelihood-differences-for-aligning-language-models-with-a-target-model",
  "paper_ref": "domain-mixture-design-via-log-likelihood-differences-for-aligning-language-models-with-a-target-model",
  "topic_slug": null,
  "benchmark_ref": null,
  "dataset_ref": null
}