Skip to content
K = 0.731
Benchmark sources

EQ-Bench 4 · Emotional and interpersonal understanding

EQ-Bench · Understanding emotions, implicit needs and different communication preferences in multi-turn conversations.
Official benchmark
Across e/accUnder observation
Evidence budgetNot scored
Upstream data timeUnverified
Last successful syncCollection not started

What it measures and how

120 simulated characters, 16 role-play turns and rotating blind judgments by three model judges. Use only Elo; behavioral traits are not capabilities where higher is always better. The older v3 adds no duplicate votes.

How this evidence is used

Under observation: public data and licensing are available, but a new evaluation round is pending. Validity windows are not extended to meet category source counts.

Limitations and attribution

Simulated English conversations have not been validated by human experts. Results do not equal real user satisfaction or Chinese emotional intelligence. Current results exceed the 45-day window.

Data license: MIT · Metadata declaration in the EQ-bench-site README

Results are published by EQ-Bench. Raw scores and the e/acc consensus index use different scales and cannot be added directly.