ResearchBenchmarksQuestion Answering on LongFormQA max len T=1024Follow63F1 ScoreVanilla62.37662.53862.762.862May 28, 2026Evaluation ResultsMethodMethodLinksF1 ScoreToxicity RateVanillaBackbone=LLaMA-2-13BBackbone=LLaMA-2-13B2026.05636COFTBackbone=LLaMA-2-13BBackbone=LLaMA-2-13B2026.0562.42.5