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

Attribution Faithfulness Evaluation on TwinMarket (test)

7,790Risk Drop

Ours (Shapley-based framework)

-1,238.241,105.633,449.55,793.37Jan 28, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
7,790
2026.01
6,514
2026.01
6,356
2026.01
6,005
2026.01
5,646
2026.01
5,554
2026.01
5,282
2026.01
5,088
2026.01
4,621
2026.01
4,606
2026.01
4,491
2026.01
4,112
2026.01
4,072
2026.01
3,820
2026.01
3,820
2026.01
2,550
2026.01
2,176
2026.01
1,700
2026.01
1,663
2026.01
1,663
2026.01
836
2026.01
476
2026.01
464
2026.01
277
2026.01
203
2026.01
139
2026.01
115
2026.01
100
2026.01
79
2026.01
62
2026.01
57
2026.01
55
2026.01
49
2026.01
49
2026.01
43
2026.01
27
2026.01
-14
2026.01
-14
2026.01
-17
2026.01
-27
2026.01
-39
2026.01
-51
2026.01
-117
2026.01
-134
2026.01
-134
2026.01
-143
2026.01
-143
2026.01
-338
2026.01
-494
2026.01
-891