TELOS Governance Benchmark Dataset: Out-of-Scope Detection for Task-Oriented Dialogue Systems

Benchmark results demonstrating the TELOS governance layer for purpose-aligned dialogue systems.

TELOS provides continuous fidelity measurement of user-to-Ai alignment against an

enterprise-declared Primacy Attractor (PA) to detect out-of-scope queries and conversational drift

without requiring model fine-tuning.

Key Results:

- CLINC150 OOS Detection: 0% → 78% improvement with governance layer

- MultiWOZ Drift Detection: 100% detection across all categories

- Adversarial Resistance: Jailbreak attempts produce negative fidelity scores

Benchmarks:

- CLINC150 (Larson et al., 2019): Intent classification with out-of-scope detection

- MultiWOZ 2.4 (Ye et al., 2022): Multi-domain task-oriented dialogue with injected drift

Validation Status:

Proof-of-concept validation on academic benchmarks. Not certified for production deployment without

domain-specific calibration.

License:

Apache 2.0