CoT Oracle Evals - a ceselder Collection

ceselder 's Collections

CoT Oracle Training Data

CoT Oracle Evals

CoT Oracle Evals

updated 10 days ago

Eval datasets for the CoT Trajectory Oracle — detecting unfaithful chain-of-thought reasoning via activation trajectories.

ceselder/cot-oracle-eval-decorative-cot

Viewer • Updated 16 days ago • 56 • 50
ceselder/cot-oracle-eval-rot13-reconstruction

Viewer • Updated 16 days ago • 100 • 50
ceselder/cot-oracle-truthfulqa-hint-admission-unverbalized

Viewer • Updated 14 days ago • 11k • 13
ceselder/cot-oracle-truthfulqa-hint-admission-verbalized

Viewer • Updated 14 days ago • 4.38k • 13