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

Long Context Understanding on LongBench (Comprehensive Subtask Metrics)

27.77NQA Score

Full

20.354822.279924.20526.1301Apr 9, 2026
Updated 9d ago

Evaluation Results

MethodLinks
2026.04
27.7744.5849.5360.9348.936.0233.2923.5924.8992.2510069.7566.6352.16
2026.04
27.2436.7156.1556.0344.2830.3732.5522.9626.6791.849968.2265.5550.58
2026.04
26.836.1756.7457.8643.8831.0432.8723.426.8792.079969.7964.2850.83
2026.04
26.6835.5355.6860.0345.9130.6632.7723.0426.5691.0998.568.3364.7550.73
2026.04
26.4444.0349.9361.3545.5227.2533.8324.0624.9392.7510069.8364.451.1
2026.04
25.6144.1753.453.4838.2932.1433.1723.5324.9390.7110067.5664.9250.14
2026.04
25.0743.9549.8760.9747.8635.1633.3424.1624.8792.2510070.2665.6551.8
2026.04
25.0133.7752.2144.9936.8323.0331.2921.8726.2989.299665.461.7746.75
2026.04
24.7744.3552.6753.5938.0231.133.5823.6724.7289.8899.567.6765.1749.9
2026.04
23.1543.0147.8854.9344.5632.4332.0322.6824.8390.4199.567.5358.3149.33
2026.04
22.544.7654.0554.2236.2633.1333.5123.4724.6590.0610068.8565.9450.1
2026.04
20.6440.1351.0145.7438.4627.2532.1322.2124.9587.5598.566.8661.9547.49