All 34 questions from ARC public eval set that o3 high failed:

Task 05a7bcf2

Task 0934a4d8

Task 09c534e7

Task 0d87d2a6

Task 16b78196

Task 1acc24af

Task 212895b5

Task 25094a63

Task 256b0a75

Task 3ed85e70

Task 40f6cd08

Task 47996f11

Task 4b6b68e5

Task 52fd389e

Task 79fb03f4

Task 891232d6

Task 896d5239

Task 8b28cd80

Task 93c31fbe

Task a3f84088

Task aa4ec2a5

Task ac0c5833

Task b457fec5

Task b7999b51

Task b9630600

Task c6e1b8da

Task d931c21c

Task d94c3b52

Task da515329

Task e1d2900e

Task e619ca6e

Task e681b708

Task f3b10344

Task f9d67f8b