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

RealTalk

Benchmarks

Task NameDataset NameSOTA ResultTrend
Long-term Conversation Question AnsweringRealtalk
Multi-hop Score49.2
84
Interactive Head GenerationRealTalk 1.0 (test)
SID1.083
10
Episodic RecollectionREALTALK
F1 Score26.2
8
Long-term Memory Question AnsweringREALTALK (Overall)
Multi-hop F137.41
7
Facial Expression GenerationRealTalk
Variation0.0234
7
Listener Response GenerationRealtalk 1.0 (user study)
Appropriateness Score4.5
4
Head Pose GenerationRealTalk (test)
FD3.158
4
Facial Expression GenerationRealTalk (test)
FD0.069
4
Listening Head GenerationRealTalk (Dtest)
FD-exp17.63
4
Image Quality AssessmentRealTalk (test)
SSIM0.58
4
Listening head generationRealTalk (test)
E-Consis Score3.798
3
Showing 11 of 11 rows