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

Multi-turn instruction following on SysBench

92.31CSR

ImpRIF-32B

56.627665.891375.15584.4187Sep 30, 2025Oct 21, 2025Nov 11, 2025Dec 2, 2025Dec 23, 2025Jan 13, 2026Feb 4, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.02
92.3186.0469.2
2026.02
91.383.665.4
2026.02
88.4678.857.6
2026.02
88.4179.0859.6
2026.02
87.7576.8455.6
2026.02
86.6875.0853.2
2026.02
86.2475.6453.76
2026.02
86.1973.6850.4
2026.02
84.6671.4847.2
2026.02
82.7570.6846
2026.02
81.9866.5242
2026.02
81.8267.0841.64
2026.02
81.2266.8444
2026.02
80.1263.842.4
2026.02
77.7260.235.2
2026.02
77.161.237.8
2026.02
75.8758.4836.4
2025.09
74.358.233.9
2025.09
73.756.432.7
2025.09
73.357.332.2
2025.09
73.257.836.5
2025.09
7357.936
2025.09
7357.936
2025.09
72.258.832.3
2025.09
71.15329.9
2025.09
69.852.433.3
2025.09
69.352.233.2
2025.09
6951.632.2
2025.09
68.550.431.8
2025.09
67.850.127.2
2025.09
67.647.927.1
2025.09
67.346.925.1
2025.09
67.148.426
2025.09
66.846.425.2
2025.09
65.749.225.2
2025.09
65.146.524.2
2025.09
64.844.824.7
2025.09
5840.620.2