Report
Protein-folding post-training reportedly improves reasoning benchmark scores
HuggingPapers says FoldingCorpus post-training raised macro-average accuracy from 45.09% to 48.33% across 10 reasoning benchmarks.
TLDR
HuggingPapers says post-training on FoldingCorpus raised macro-average accuracy from 45.09% to 48.33% across 10 reasoning benchmarks. The post also claims 2.7–3.5× better structure prediction.
Combined views
5K
2 Sources, first seen ago
