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

Retrieval on Needle-in-a-Haystack L=8k

100Accuracy

Baseline SFT

55.716867.213478.7190.2066Mar 15, 2026Mar 20, 2026Mar 25, 2026Mar 30, 2026Apr 4, 2026Apr 9, 2026Apr 15, 2026
Updated 3d ago

Evaluation Results

MethodLinks
100
2026.04
100
2026.04
100
2026.04
100
2026.04
100
2026.03
97.56
2026.03
94.38
2026.03
93.21
2026.03
91.02
2026.03
90.38
2026.03
86.52
2026.03
85.92
2026.03
82.67
2026.03
81.78
2026.03
81.62
2026.03
78.96
2026.03
61.63
2026.03
57.42