List your company

ResearchFormerly WizWand

SAGE

Benchmarks

TaskDatasetSOTA resultTrendResults

Safety evaluation

SAGE-Eval

90Safety

18

Emotional Support Conversation

SAGE (test)

85.07Sentience

14

LLM-as-a-Judge Robustness

Sage (Hard)

55.9Factuality (IPI)

13

LLM-as-a-Judge Robustness

Sage Easy

0.059Factuality Error (IPI)

13

Emotional Support Dialogue

SAGE

72.1Average Score

12

Open-Ended Question Answering

SAGE Web Search

35.1Weighted Recall (Com. Sci.)

12

Short-Form Question Answering

SAGE Web Search

63.3Accuracy (Com. Sci.)

12

Cultural Reasoning

SAGE (test)

97.57Accuracy

10

Support-only Emotional Intelligence Evaluation

SAGE

39.6Support Score

4

Multi-hop Question Answering

SAGE Small-scale (evaluation)

4.9# Search

1