Share your thoughts, 1 month free Claude Pro on usSee more

Long-context Instruction Following on LIFBench

64.4List Score

DIRECTER

Updated 1mo ago

Evaluation Results

Method	Links
DIRECTER 2026.03		64.4	70	51.7
*-marked 2026.03		64.3	66.9	44.9
Zero-shot 2026.03		63.4	68.6	40.9
"-marked 2026.03		63.4	69.9	41
PASTA* 2026.03		61.8	66	47.8
SpotLight* 2026.03		61.4	70.8	38.8
PASTA 2026.03		61.1	62.8	22.5
Few-shot 2026.03		55.5	57.7	42.2
SpotLight 2026.03		55.2	56.3	36.8