@teortaxesTex it behaves much more like a subfrontier chinese model in this tendency. at a glance, the worst cases seem to be from 4.8 the prior override trigger is just, too strong when the boundaries between "original text to preserve" & "original generative content" are closely interleaved
@teortaxesTex exact precision seems esp. brutal on recent opuses (after the 4.7 continued pretrain & tokenizer switch) newer opuses just can't help themselves. they like to paraphrase/'correct'/embellish the parts of the ground truth that are supposed to remain stable