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

Long-context language evaluation on LongBench

48.43SQA

Baseline

23.36629.87336.3842.887Dec 18, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.12
48.4343.1825.9963.2234.8359.8945.92
2025.12
48.2243.0525.9763.2234.7259.3345.75
2025.12
48.1542.7125.3863.0534.6759.1645.52
2025.12
47.8340.525.3463.0934.562.9545.7
2025.12
47.5641.3524.1564.3234.8365.946.35
2025.12
47.4139.8425.2161.3233.6762.745.02
2025.12
46.9742.8225.7162.3334.1454.3644.39
2025.12
44.8742.3423.7461.9934.562.7145.02
2025.12
44.7140.4624.3462.7233.9457.5343.95
2025.12
44.1940.3823.0260.833563.9844.57
2025.12
40.8738.4723.0362.3834.6763.2143.77
2025.12
24.8325.0522.4156.331259.8933.42
2025.12
24.3328.4421.0156.8212.564.3234.57