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

Long-context Retrieval on Needle-in-a-Haystack

100Retrieval Accuracy

Qwen2.5-3B-Instruct (Teacher)

-4235077Mar 29, 2025Jun 8, 2025Aug 19, 2025Oct 29, 2025Jan 9, 2026Mar 21, 2026Jun 1, 2026
Updated 19h ago

Evaluation Results

MethodLinks
2025.12
100
2025.12
100
2025.12
100
2025.12
100
2025.12
100
2025.03
100
2026.06
100
2026.06
100
2026.06
100
2026.06
100
2026.06
100
2026.06
100
2026.06
100
2026.06
100
2026.06
100
2025.12
99.8
2025.12
99.8
2026.06
99.5
2025.12
99.4
2026.06
99
2026.06
96.5
2025.12
95.4
2026.06
86.5
2026.06
82.5
2025.12
68.4
2026.06
63.5
2025.03
50
2025.03
10
2026.06
0