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

Long-context generation on LongBench

23.94QMSum

baseline

-0.8645.575512.01518.4545Jun 3, 2024Sep 18, 2024Jan 3, 2025Apr 21, 2025Aug 6, 2025Nov 21, 2025Mar 9, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
23.94----42.98-64.42-
2026.03
23.43----46.26-66.56-
2026.03
22.97----44.27-63.34-
2026.03
22.25----42.47-60.78-
2026.03
21.53----40.63-60.18-
2024.06
21.2424.4126.536886.8141.9727.5743.0842.45
2024.06
21.1519.9825.856478.9142.2423.1547.6640.37
2024.06
21.0723.2726.916682.5941.0625.5348.2341.83
2024.06
20.7218.9326.5966.583.0442.6726.0238.0940.32
2024.06
20.2417.9724.65867.237.9419.4129.3434.34
2024.06
20.2317.6723.395980.7538.7221.7937.3137.36
2026.03
19.31----42.14-53.67-
2026.03
19.08----40.85-54.78-
2026.03
15.83----36.42-35.24-
2026.03
14.92----33.87-41.99-
2026.03
13.29----34.2-44.93-
2026.03
7.28----25.81-47.03-
2024.06
3.931.622.6410.810.611.8714.973.43
2024.06
2.952.183.541.51.830.356.7111.573.83
2024.06
24.116.05151.621.554.2425.927.56
2024.06
1.780.682.8391.130.4513.838.464.77
2026.03
0.09----1.98-5.65-
2026.01
--------45
2026.01
--------44.7
2026.01
--------47
2026.01
--------46.7
2026.01
--------48.5
2026.01
--------48.2