Benchmark RadarAI BENCHMARK PROFILE
Hy-MultiTurn
Hy-MultiTurn is a Chinese benchmark for deep multi-turn dialogue understanding, with six controlled evaluation modes covering constraint memory, object localization, and action suppression across 12-76 turn dialogues.
- Released
- 2026-07-31
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Evaluates long multi-turn dialogue capabilities that existing benchmarks miss, offering controlled modes for diagnosing model failures.
Motivation
Long-running multi-turn interactions with chatbots and agents are now common, and a correct response often depends on remembering earlier details, tracking later revisions, identifying intended objects or referents, and withholding action when required conditions are unmet.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.