AnoMod
Source
- Dataset metadata snapshot: anomod-2026
- arXiv: https://arxiv.org/abs/2601.22881
- Official dataset: https://zenodo.org/records/18342898
- Official code: https://github.com/EvoTestOps/AnoMod
Core Claim
AnoMod is a multimodal anomaly-detection and root-cause-analysis dataset for microservice systems. It adds API responses and code coverage reports to logs, metrics, and distributed traces.
Dataset Notes
- AnoMod is built on SocialNetwork and TrainTicket.
- SocialNetwork has 21 services; TrainTicket has 41 microservices.
- It defines 24 anomaly cases across performance, service, database, and code levels.
- TrainTicket contains 444.6K log lines, 63975 traces, 33 collected metrics, and 98073 API requests.
- SocialNetwork contains 3958.5K log lines, 2635 traces, 29 collected metrics, and 1950 API requests.
Why It Matters
AnoMod is the strongest dataset here for testing multimodal fusion beyond standard observability telemetry. API responses connect telemetry to user-visible symptoms, and code coverage can connect symptoms back to executed code regions.
Gotchas
- The dataset is centered on anomaly detection and RCA, not forecasting.
- It is not a simple service-graph tensor; dependency structure is available through traces and system architecture.
- Zenodo lists the dataset as CC-BY-4.0, while the GitHub scripts use MIT.
Foundation TSFM Relevance
| Agenda slot | Verdict | Evidence | Missing pieces |
|---|---|---|---|
| Benchmarks: what level of modeling is tested? | partially closes | Provides controlled microservice anomaly and RCA tasks over logs, metrics, traces, API responses, and code coverage. | Does not test forecasting, counterfactual rollout, or action choice quality. |
| Context interface | adjacent | Traces, system dependencies, API responses, and code coverage expose system context beyond raw metric values. | Needs a standardized context schema for topology, deployments, and operator events. |
| Causal structure, counterfactuals, and control | warning | Anomaly injections are controlled benchmark events. | They are not logged remediation actions or candidate interventions for counterfactual planning. |