Benchmark RadarAI BENCHMARK PROFILE
HeraBench
HeraBench is a fault-injected benchmark for multi-device agent workflows on Linux and Android, evaluating hierarchical replanning under injected strategy- and device-level failures.
- Released
- 2026-06-18
- Readiness
- Paper only
- Primary field
- Consumer & Productivity
Why it matters
Current multi-device agent benchmarks lack systematic fault injection to test recovery capabilities, making it difficult to compare hierarchical versus global replanning approaches.
Motivation
Real-world computer-use tasks often span multiple applications and devices, requiring agents to coordinate heterogeneous environments under dynamic runtime failures.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.