TelecomTS: A Multi-Modal Observability Dataset for Time Series and Language Analysis

Researchers introduce TelecomTS, a large-scale observability dataset from 5G telecommunications networks designed to advance time series analysis and anomaly detection. The dataset addresses a critical gap in AI research by providing de-anonymized, scale-preserved metrics that reflect real-world system monitoring challenges, while benchmarking reveals that current foundation models struggle with the noisy, high-variance characteristics of enterprise observability data.

TelecomTS represents a significant contribution to the AI research community by tackling a persistent problem: the scarcity of public, production-grade observability datasets. Enterprise monitoring systems generate massive volumes of time series data, yet these remain largely proprietary due to privacy and competitive concerns. This new dataset breaks that barrier by providing real 5G network metrics with preserved absolute scale information—a critical detail typically stripped during anonymization that severely limits model utility for practical applications like anomaly detection and root cause analysis.

The research identifies a fundamental mismatch between conventional time series benchmarks and actual observability data characteristics. Unlike climate or financial datasets, observability metrics are zero-inflated, highly stochastic, and exhibit minimal temporal structure. These properties render many existing state-of-the-art models ineffective, creating opportunities for specialized foundation model development. The benchmark results underscore that scaling metrics appropriately matters significantly for real-world performance.

For the AI and enterprise software sectors, this dataset accelerates development of specialized observability tools. Startups and vendors building monitoring solutions gain access to realistic evaluation benchmarks, while researchers can develop models optimized for real operational challenges rather than idealized scenarios. The emphasis on multi-modal reasoning—combining time series with natural language analysis—points toward the next generation of intelligent observability platforms that can automatically diagnose system issues through both metric analysis and contextual reasoning.

Looking forward, the research signals growing maturity in enterprise AI applications. As observability becomes increasingly critical for complex distributed systems, datasets like TelecomTS will likely spawn new model architectures and commercial tools designed specifically for operational intelligence.

• →TelecomTS fills a critical gap by providing the first large-scale, publicly available observability dataset from production 5G networks with preserved absolute scale information.
• →Current state-of-the-art time series and foundation models underperform on real observability data due to its zero-inflated, highly stochastic nature distinct from academic benchmarks.
• →Preserving absolute scale information in metrics significantly improves model performance for practical anomaly detection and root cause analysis tasks.
• →The dataset enables multi-modal reasoning combining time series analysis with natural language understanding for automated system diagnostics.
• →This research creates opportunities for specialized foundation models optimized for enterprise observability rather than generic time series tasks.