Benchmark RadarAI BENCHMARK PROFILE
CallBench
CallBench is a Chinese benchmark for evaluating dual-goal coordination in phone call assistants, with 50,000 multi-turn dialogues across six scenarios and a preset-aware turn-level scoring protocol.
- Released
- 2026-06-22
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing dialogue benchmarks focus on single explicit goals, while real phone assistants must balance the owner's preset and the caller's dynamic goal. CallBench provides a reusable evaluation to measure turn-level decisions under proxy constraints.
Motivation
Target-oriented dialogue systems have demonstrated strong capabilities in completing user goals through interactive conversations.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.