Fix over-stated quality claims: model is first-sentence-coherent then loops, not "fluent/on-topic"
#9
by Compactbot - opened
No description provided.
Merging: the card over-claimed quality ("fluent, grammatical, on-topic" / "genuinely coherent, on-topic output"). Running the actual published weights shows greedy output is coherent for ~the first sentence then collapses into repetition loops, and sampled output drifts incoherent within a few sentences. This correction makes the card honest.
Compactbot changed pull request status to merged