Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

SGD

Benchmarks

Task NameDataset NameSOTA ResultTrend
Dialog State TrackingSGD 15 tasks CL
Avg JGA76.3
23
Natural Language GenerationSGD (test)
BLEU28.6
18
User Satisfaction EstimationSGD
Accuracy64.8
14
Task-oriented DialogueFewShotSGD unseen schemata (test)
BLEU28.76
13
Task-oriented DialogueFewShotSGD seen schemata (test)
BLEU29.28
13
Dialogue State TrackingSGD (test)
JGA86.5
11
Dialog Structure InductionSGD (test)
Purity46.8
9
Dialogue State TrackingSGD
JGA (Payment)24.7
8
User Satisfaction EstimationSGD 5% training size (test)
Precision75.3
8
Task-Oriented DialogueSGD 1.0 (test)
Inform Rate81.29
6
Structure InductionSGD Real (test)
AMI0.559
6
Hidden Representation LearningSGD Real (test)
Class-Balanced Acc (Full)66.3
6
Dialogue State TrackingSGD-X v1-v5 variants (test)
Joint Goal Acc (Original)86.4
6
Goal completionSGD (test)
Inform Rate50.4
5
Dialogue State TrackingSGD to MultiWoz (test)
Average JGA51.2
5
Dialogue State TrackingSGD All Domains (test)
Joint GA32.1
4
Dialogue State TrackingSGD Unseen Domains (test)
Joint GA24.4
4
Natural Language GenerationSGD (Overall)
Naturalness2.46
4
Natural Language GenerationSGD Seen domains
Naturalness2.48
4
Natural Language GenerationSGD Unseen domains
Naturalness2.46
4
Natural Language GenerationSGD 1.0 (overall)
BLEU28.6
4
Natural Language GenerationSGD 1.0 (unseen domains)
BLEU Score22.2
4
Natural Language GenerationSGD seen domains 1.0
BLEU29.4
4
Dialog Structure InductionSGD Synthetic (test)
Purity81
3
Action Selection Task (AST)SGD (out-of-distribution)
B-Slot Acc61.3
3
Showing 25 of 32 rows