pith. sign in

Chunky post-training: Data driven failures of generalization

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.AI 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

How Well Do Models Follow Their Constitutions?

cs.AI · 2026-05-22 · unverdicted · novelty 6.0

Newer Claude and GPT models violate their labs' constitutions at substantially lower rates (15% to 2% and 11.7% to 3.6%) than prior generations, though causes cannot be isolated and some failure modes persist.

citing papers explorer

Showing 1 of 1 citing paper.

  • How Well Do Models Follow Their Constitutions? cs.AI · 2026-05-22 · unverdicted · none · ref 11

    Newer Claude and GPT models violate their labs' constitutions at substantially lower rates (15% to 2% and 11.7% to 3.6%) than prior generations, though causes cannot be isolated and some failure modes persist.