List your company

ResearchFormerly WizWand

Multi-turn ambiguous query clarification

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

ClarifyMT-Bench Average 1.0 (full dataset)

Llama-3.1-70B-It

77.1Accuracy

10

Feb 26, 2026

ClarifyMT-Bench Refusal persona 1.0 (full dataset)

72.9Accuracy

10

Feb 26, 2026

ClarifyMT-Bench Factually-Wrong persona 1.0

Claude-Sonnet-4.5

93.7Accuracy

10

Feb 26, 2026

ClarifyMT-Bench Contradictory persona 1.0

Gemini-2.5-Flash

97.4Accuracy

10

Feb 26, 2026

ClarifyMT-Bench Off-Focus persona 1.0

Claude-Sonnet-4.5

96.1Accuracy

10

Feb 26, 2026

ClarifyMT-Bench Partial-Vague persona 1.0

Gemini-2.5-Flash

95.5Accuracy

10

Feb 26, 2026

ClarifyMT-Bench Precise persona 1.0

96.6Accuracy

10

Feb 26, 2026