zsRE
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ZsRE N = 2000
100Reliability
12
zsRE
40Top-1 Accuracy
12
ZsRE 1000 edits Sequential (test)
100Efficiency
12
ZsRE 500 edits Sequential (test)
100Efficiency
12
ZsRE 100 edits Sequential (test)
100Efficiency
12
zsRE KILT (test)
72.55KILT Accuracy
12
ZsRE (T = 10000)
96.6Efficacy
11
Bi-ZsRE
80.53Efficacy
10
ZSRE Continual
63.28Averaged Soft Token Accuracy
10
ZSRE (Original)
37.54Averaged Soft Token Accuracy
10
ZsRE massive editing (2,000 items)
97.6Efficacy Success
9
zsRE
25Error Rate (ERR)
9
zsRE (test)
34Top-1 Accuracy
9
ZsRE (test)
99.5Efficacy Success Rate
8
ZsRE
93.7Efficacy Success
8
ZsRE 3,000 samples
77.8Rel Score (QA Context)
8
ZsRE (val)
1Rel.
8
zsRE (val)
99.8Efficacy
8
zsRE
72TRR
7
ZsRE No Context 10K
31.1Reliability
6
zsRE
100Attack Performance
6
zsRE
99.6Efficacy
6
ZsRE N=10000 edits (test)
99Reliability
5
ZsRE N=5000 edits (test)
99Reliability
5
ZsRE N=2000 edits (test)
99Reliability
5