Instruction Following
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
86.5Prompt Strict Accuracy
22
May 8, 2026
47.67Average Score
21
May 20, 2026
90Strict Accuracy
21
May 18, 2026
74Improvement Score
21
Jun 15, 2026
84.63IFScore
21
May 15, 2026
34.69Prompt-level Accuracy
21
Apr 6, 2026
86.7PLA
21
Mar 18, 2026
3.15SR Score
21
Apr 30, 2026
119.4Pairwise Score (ChatGPT vs Sys)
21
Feb 26, 2026
39Rouge-L
20
Jun 30, 2026
36.1Rouge-L
20
Jun 30, 2026
19.7Rouge-L
20
Jun 30, 2026
23.2Rouge-L
20
Jun 30, 2026
90.65I-Following Score
20
Jun 19, 2026
51.53DROP Score
20
Apr 21, 2026
64.31MMLU Accuracy
20
Apr 21, 2026
68.9Win Rate (%)
20
Feb 26, 2026
83Win Rate
20
Feb 26, 2026
64.2Win Rate
20
Feb 26, 2026
77.8Win Rate
20
Feb 26, 2026
65.2Win Rate
20
Feb 26, 2026
70.4Win Rate
20
Feb 26, 2026
57.5Instruction Following Score
19
May 29, 2026
75.4IFBench Score
19
Mar 27, 2026
59.5A Wins
19
Feb 26, 2026