List your company

ResearchFormerly WizWand

RealTalk

Benchmarks

TaskDatasetSOTA resultTrendResults

Long-term Conversation Question Answering

Realtalk

49.2Multi-hop Score

84

Multi-Turn Dialogue Generation

REALTALK zero-shot

9.789PPL

14

Multi-turn Dialogue Generation

REALTALK (per-session)

9.754PPL

12

Information Retrieval

REALTALK

62.6Recall@10

10

Interactive Head Generation

RealTalk 1.0 (test)

1.083SID

10

Episodic Recollection

REALTALK

26.2F1 Score

8

Multi-session Dialogue Generation

REALTALK two-session (test)

9.87PPL

7

Long-term Memory Question Answering

REALTALK (Overall)

37.41Multi-hop F1

7

Facial Expression Generation

RealTalk

0.023Variation

7

Listener Response Generation

Realtalk 1.0 (user study)

4.5Appropriateness Score

4

Head Pose Generation

RealTalk (test)

3.158FD

4

Facial Expression Generation

RealTalk (test)

0.069FD

4

Listening Head Generation

RealTalk (Dtest)

17.63FD-exp

4

Image Quality Assessment

RealTalk (test)

0.58SSIM

4

Listening head generation

RealTalk (test)

3.798E-Consis Score

3