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

Long-context multi-task evaluation on LongBench-E

45.25Qasper

FluxAttn

33.70636.70339.742.697Apr 8, 2026
Updated 8d ago

Evaluation Results

MethodLinks
2026.04
45.2554.4254.5441.3434.5426.1668.3391.9142.17997.6747.7465.3552.2851
2026.04
45.154.656.6741.8834.0925.5171.3390.9342.6310.6294.6743.6459.3551.6750
2026.04
44.7754.1257.3543.4334.3125.872.3391.3242.629.3398.3345.4860.752.4262
2026.04
44.0653.4459.6244.0834.526.027190.5442.9412.6799.3347.7863.8553.28-
2026.04
43.7653.4157.3639.4332.9625.6370.3391.2742.21198.6745.666.1752.351
2026.04
42.1453.1358.5343.534.6626.0670.6791.4643.13899.6750.9164.7853.0772
2026.04
41.7851.5555.9641.733.2423.3469.3389.3541.620.598.9357.5469.3952.1350
2026.04
41.5152.3650.4637.5734.2525.8666.3389.9341.729.079756.8466.2351.6850
2026.04
41.2249.9258.9844.2133.2723.4271.3386.7741.83298.3356.0866.3152.16-
2026.04
4149.7658.1944.3633.3223.357088.7741.71.3399.6755.667.2252.2247
2026.04
40.8251.3157.5744.5133.3223.3571.3386.7341.79294.3355.0465.8951.6550
2026.04
40.4150.2657.7840.5733.2723.5169.6787.1942.121.3399.3355.4165.5151.5751
2026.04
40.350.4956.0240.933.0123.5571.6788.3141.610.3310059.4668.2752.1846
2026.04
39.9250.0455.7240.8133.0323.57288.4840.960.3399.2258.5769.4652.0546
2026.04
37.9551.251.9436.4833.1123.366987.942.1119857.0567.6650.850
2026.04
35.8349.8447.0932.2433.3223.769.3385.8739.754.594.5750.5657.4348.2250
2026.04
35.7451.745.8332.3433.5723.666987.2339.813.593.7450.8159.2848.3253
2026.04
35.5552.0245.3731.7633.3223.76988.239.743.8391.5148.5856.3847.7250
2026.04
35.2152.1644.8132.1533.4723.4570.6788.2239.742.3396.8450.8457.9348.45-
2026.04
35.151.6849.6532.8633.0423.4269.3388401.6794.4751.458.6848.5944
2026.04
35.0249.4449.6432.2733.2623.4869.3388.2939.781.594.5653.4459.6948.7244
2026.04
35.0250.8945.1734.2433.0223.536988.0840.383.9496.0651.686048.7647
2026.04
34.6350.7449.736.4134.2525.787091.4542.139.897.3353.5968.5552.1150
2026.04
34.1550.7844.4832.8932.9623.5367.6788.6939.553.1790.174954.0747.1650