The DOI matches its Crossref record; 38 of the 38 references it cites were found in a registry or an index; 15 of the 39 sources its claims cite were checked for support.
Digital Twins at the Edge: A High-Availability Framework for Resilient Data Processing in IoT Sensor Networks
Open accessAbstract
This paper introduces the Operational Digital Twin, a live state-synchronized standby system for node-level failover in resource-constrained edge environments.
Overview
CoChat Summary
TL;DR
The paper introduces the Operational Digital Twin (ODT), a state-synchronized standby edge node that assumes the primary node's role on failure, and validates it on a two-node Raspberry Pi 5 testbed. Fault-injection experiments show ~1.35 s VIP failover, 100% success across 500 cycles, and 10.81% message loss during transitions.
Key findings
- Mean VIP failover latency was 1.353 s (200 tests) and 1.335 s (500 tests), with 95% of failovers completed within 2.5 s; 100% success rate across 500 consecutive cycles with no degradation.
- Data-processing service recovered in a mean of 0.271 s (200 tests), while MQTT broker recovery was the bottleneck and highly variable (range 0.012–7.630 s, mean 1.339 s).
- Average message loss during failover was 10.81% (delivery rate 89.19%), consistent with the 1.345 s vulnerability window at a 10 Hz publication rate.
- Resource overhead was very low: CPU below 3% and memory below 10% during a 6 h campaign; split-brain was detected in ≈6 s and resolved in ≈12 s via an external witness device.
Methods
- Design: a three-layer architecture (ESP8266 sensing layer, primary–backup Raspberry Pi 5 edge layer, Virtual IP virtualization layer) using MQTT/Mosquitto, keepalived/VRRP for VIP failover, rsync over SSH for 30 s configuration synchronization, and a custom vip-monitor service.
- Data/sample: two Raspberry Pi 5 nodes (8 GB RAM) plus ESP8266 NodeMCU sensors (DHT22, MQ-7) on a dual-band router; baseline round-trip latency below 2 ms; an additional Raspberry Pi 5 served as a witness for partition tests.
- Analysis: automated fault injection (stopping keepalived or applying iptables rules) across campaigns of 10, 200, and 500 cycles, measuring failover latency, service-recovery time, reliability, MQTT throughput/latency/loss, resource utilization, and partition tolerance, plus sensitivity analysis over payload size, test duration, and partition length.
Limitations
- Stated: the two-node homogeneous testbed does not reflect real-world heterogeneity or complexity; scaling to larger clusters would introduce leader-election contention, quorum logic, state-propagation scalability, and cascading failures; hardware heterogeneity may create capacity asymmetries; network conditions in the lab are less variable than operational settings; the 10 Hz periodic workload does not capture bursty or stateful IoT traffic; and six hours/500 cycles cannot reveal long-term drift, resource exhaustion, or hardware degradation.
- Evident from the text: the 10.81% message loss requires application-layer mitigation (buffering, retries, QoS), and the consistency-over-availability partition policy introduces a deliberate window of service unavailability; the state-of-the-art comparison (Table 12) draws on independent testbeds with different hardware and workloads, which the authors themselves caution against comparing directly.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
13 Valid · 1 Not valid · 9 Not checked · 96 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Second, at an implementation level, existing resilience solutions for the edge are either too heavyweight (container orchestration requiring significant resources [8,35]), too slow (cold standby with recovery times exceeding minutes [11,16]), or too narrow (sensor-level backups that cannot assume full node functionality [13]).”
§ 2. Related Work > 2.3. Positioning and Research Gap
What the sources say
- Not checked — no full text of the sourcecites [8] Barletta, M., De Simone, L., Corte, R.D., and Di Martino, C. (2024, January 8–11). Failover Timing Analysis in Orchestrating Container-based Critical Applications. Proceedings of the 2024 19th European Dependable Computing Conference (EDCC), Leuven, Belgium. source
- Partly supportedcites [35] Yakubov, D., and Hästbacka, D. (2025). Comparative Analysis of Lightweight Kubernetes Distributions for Edge Computing: Security, Resilience and Maintainability. Proceedings of the Service-Oriented and Cloud Computing: 11th IFIP WG 6.12 European Conference, ESOCC 2025, Bolzano, Italy, 20–21 February 2025, Springer. sourceExperimental — accuracy not yet measured
The Conclusion (p. 7) says standard Kubernetes may be resource-intensive for some edge deployments, and that the resilience benefits of KubeEdge and OpenYurt come with increased resource consumption. But it also describes k3s and k0s as suitable for resource-constrained edge settings, so the paper does not support treating edge container-orchestration resilience solutions as uniformly too heavyweight. (LLM, conf=0.84)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“While model-based DTs have been extensively studied for forecasting, anomaly detection, and what-if analysis [3,4,14], the notion of a twin that exists not to model the system but to become the system upon failure remains unexplored.”
§ 2. Related Work > 2.3. Positioning and Research Gap
What the sources say
- Supportedcites [3] Rasheed, 2020, Digital Twin: Values, Challenges and Enablers From a Modeling Perspective, IEEE Access source
The citation is attached to the statement about studied uses of model-based digital twins, not to the citing authors’ assertion that a failure-replacement twin remains unexplored. The source discusses prediction and what-if analysis in “Machine Learning and Artificial Intelligence,” anomaly detection in the same section, and scenario assessment in “Value of Digital Twin” (page numbers unavailable). (LLM, conf=0.90)
- Not checked — no full text of the sourcecites [4] Huang, 2021, Digital Twin-driven Online Anomaly Detection for an Automation System Based on Edge Intelligence, J. Manuf. Syst. source
- Not checked — no full text of the sourcecites [14] Kim, 2025, Digital Twin-empowered Intelligent Computation Offloading for Edge Computing in the Era of 5G and Beyond: A State-of-the-Art Survey, ICT Express source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Second, at an implementation level, existing resilience solutions for the edge are either too heavyweight (container orchestration requiring significant resources [8,35]), too slow (cold standby with recovery times exceeding minutes [11,16]), or too narrow (sensor-level backups that cannot assume full node functionality [13]).”
§ 2. Related Work > 2.3. Positioning and Research Gap
What the sources say
- Could not verifycites [11] Olorunnife, K., Lee, K., and Kua, J. (2021). Automatic Failure Recovery for Container-Based IoT Edge Applications. Electronics, 10. sourceExperimental — accuracy not yet measured
The Introduction and §2.3 discuss fault tolerance and fail-over strategies, but do not describe cold standby or give its recovery time. The paper reports redeployment times exceeding two minutes for its own system, which does not establish the claim about existing cold-standby solutions (searched: cold standby, failover, recovery times). (LLM, conf=0.76)
- Partly supportedcites [16] Peniak, P., Bubeníková, E., and Kanáliková, A. (2023). Validation of High-Availability Model for Edge Devices and IIoT. Sensors, 23. sourceExperimental — accuracy not yet measured
The laboratory validation (Section 6) reports a recovery time of about 180 seconds and an application outage of more than three minutes for the spare-Edge approach. But that experiment explicitly assumes the spare Edge is active in hot-standby mode; it does not measure a cold-standby recovery time. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“These twins observe and model the system but do not themselves become the system.”
§ 5. Discussion > 5.3. The Operational Digital Twin: Positioning and Contributions
What the sources say
- Supportedcites [3] Rasheed, 2020, Digital Twin: Values, Challenges and Enablers From a Modeling Perspective, IEEE Access source
The Introduction (page unavailable) describes a digital twin as a digital representation of a physical entity that is updated with sensor data and models the asset’s behavior. The Common Challenges section distinguishes the physical asset from its digital twin, supporting the point that the twin does not literally become the system. (LLM, conf=0.89)
- Not checked — no full text of the sourcecites [4] Huang, 2021, Digital Twin-driven Online Anomaly Detection for an Automation System Based on Edge Intelligence, J. Manuf. Syst. source
- Not checked — no full text of the sourcecites [14] Kim, 2025, Digital Twin-empowered Intelligent Computation Offloading for Edge Computing in the Era of 5G and Beyond: A State-of-the-Art Survey, ICT Express source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“IoT device redeployment exceeds 70 s [11]”
§ 5. Discussion > 5.4. Comparison with State of the Art
What the sources say
- Supportedcites [11] Olorunnife, K., Lee, K., and Kua, J. (2021). Automatic Failure Recovery for Container-Based IoT Edge Applications. Electronics, 10. source
Table 1 (in the indexed Conclusions and Future Work section; page unavailable) reports an average IoT device redeployment time of 73.6 s, exceeding 70 s. This is an average, not a minimum: two of the three individual tests took less than 70 s. (LLM, conf=0.87)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
53 more claims · 2 with their cited sources checked
Create a free account to see all 119 claims and every source verdict.
- Probably not a substantial claim. 31
- It may state two separate claims as one. 35
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 100 bytes During Failover Short (1 cycle) 673.03 82.8 79.2 > 4.6. Sensitivity Analysis
“For 10 KB payloads, the difference was minimal (15.6 ms normal vs. 21.1 ms during failover in long tests)”
§ 100 bytes During Failover Short (1 cycle) 673.03 82.8 79.2
“Note: During failover, only 100-byte payloads were tested; 1 KB and 10 KB payloads were tested during normal operation only.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 4. Experimental Evaluation > 1.5 2.0 2.5 3.0 3.5 4.0
“Table 3 reveals that data processing recovers consistently within 0.3 s”
§ 0 25 50 75 100 125 150 175 200
“Data Processing Mean (s) 0.281 [0.251, 0.332] 0.271 [0.237, 0.476]”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 5. Discussion > 5.4. Comparison with State of the Art
“comparable to PureEdge (≈ 1.8 s) [20]”
§ 8 min
“≈ 1.43 s (median, AI- based)/6.06 ms”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 0 25 50 75 100 125 150 175 200
“true service continuity as perceived by applications is determined by the maximum of network and application recovery times, which in our system is dominated by MQTT broker recovery (1.345 s).”
§ 0 25 50 75 100 125 150 175 200
“Network (1.353 s) Application (1.345 s)”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 100 bytes During Failover Short (1 cycle) 673.03 82.8 79.2
“As expected, large payloads (10 KB) saturated network bandwidth, reducing through- put to approximately 182 msg/s.”
§ 100 bytes During Failover Short (1 cycle) 673.03 82.8 79.2 > 4.6. Sensitivity Analysis
“It does not reflect network bandwidth saturation, as the achieved data rate (≈14.6 Mbps) is well within the capacity of the 5 GHz Wi-Fi link used in the experiments.”
Citations
Cites
Digital twin: values, challenges and enablers from a modeling perspective
Rasheed, A., San, O., & Kvamsdal, T. (2020) · IEEE Access · doi:10.1109/access.2020.2970143 · Cited by 1,952
Containerization and the PaaS Cloud
Pahl, C. (2015) · IEEE Cloud Computing · doi:10.1109/mcc.2015.51 · Cited by 451
Edge-computing-driven Internet of Things: A Survey
Kong, L., Tan, J., Huang, J., Chen, G., Wang, S., Jin, X., Zeng, P., Khan, M., & Das, S. K. (2023) · ACM Computing Surveys · doi:10.1145/3555308 · Cited by 396
Vulnerabilities and Limitations of MQTT Protocol Used between IoT Devices
Dinculeană, D., & Cheng, X. (2019) · Applied Sciences · doi:10.3390/app9050848 · Cited by 210
Digital Twin-driven Online Anomaly Detection for an Automation System Based on Edge Intelligence
Huang, H., Yang, L., Wang, Y., Xu, X., & Lu, Y. (2021) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2021.02.010 · Cited by 183
A Survey of AI-Based Anomaly Detection in IoT and Sensor Networks
DeMedeiros, K., Hendawi, A., & Alvarez, M. (2023) · Sensors · doi:10.3390/s23031352 · Cited by 160
Wang, F., Jiang, Y., Zhang, R., Wei, A., Xie, J., & Pang, X. (2025) · Sensors · doi:10.3390/s25010190 · Cited by 119
A Review of Edge Computing Technology and Its Applications in Power Systems
Liang, S., Jin, S., & Chen, Y. (2024) · Energies · doi:10.3390/en17133230 · Cited by 84
Learning Spatiotemporal Failure Dependencies for Resilient Edge Computing Services
Aral, A., & Brandic, I. (2021) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2020.3046188 · Cited by 68
Mudassar, M., Zhai, Y., & Lejian, L. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3144026 · Cited by 59
Analysis on Cascading Reliability of Edge-Assisted Internet of Things
Fu, X., Wang, Y., Yang, Y., & Postolache, O. (2022) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2022.108463 · Cited by 48
Container Orchestration in Edge and Fog Computing Environments for Real-Time IoT Applications
Wang, Z., Goudarzi, M., Aryal, J., & Buyya, R. (2022) · Lecture notes on data engineering and communications technologies · doi:10.1007/978-981-19-3391-2_1 · Cited by 45
Tran-Dang, H., & Kim, D.-S. (2025) · ICT Express · doi:10.1016/j.icte.2025.01.002 · Cited by 38
Container Orchestration in Edge and Fog Computing Environments for Real-Time IoT Applications
Wang, Z., Goudarzi, M., Aryal, J., & Buyya, R. (2022) · arXiv.org · doi:10.48550/arxiv.2203.05161 · Cited by 36
Bhatti, D. S., Hussain, M. M., Suh, B., Ali, Z., Akobir, I., & Kim, K. (2024) · IEEE Access · doi:10.1109/access.2024.3382508 · Cited by 33
Anomaly detection using machine learning and adopted digital twin concepts in radio environments
Moharam, M. H., Hany, O., Hany, A., Mahmoud, A., Mohamed, M., & Saeed, S. (2025) · Scientific Reports · doi:10.1038/s41598-025-02759-5 · Cited by 32
DeepFT: Fault-Tolerant Edge Computing using a Self-Supervised Deep Surrogate Model
Tuli, S., Casale, G., Cherkasova, L., & Jennings, N. R. (2023) · IEEE INFOCOM 2023 - IEEE Conference on Computer Communications · doi:10.1109/infocom53939.2023.10229049 · Cited by 23
DRAGON: Decentralized Fault Tolerance in Edge Federations
Tuli, S., Casale, G., & Jennings, N. R. (2023) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2022.3199886 · Cited by 21
Low-Latency Edge-Enabled Digital Twin System for Multi-Robot Collision Avoidance and Remote Control
Mtowe, D. P., Long, L., & Kim, D. M. (2025) · Sensors · doi:10.3390/s25154666 · Cited by 19
Adaptive Data Replication in Real-Time Reliable Edge Computing for Internet of Things
Wang, C., Gill, C., & Lu, C. (2020) · 2020 IEEE/ACM Fifth International Conference on Internet-of-Things Design and Implementation (IoTDI) · doi:10.1109/iotdi49375.2020.00019 · Cited by 18
Automatic Failure Recovery for Container-Based IoT Edge Applications
Olorunnife, K., Lee, K., & Kua, J. (2021) · Electronics · doi:10.3390/electronics10233047 · Cited by 17
The Implementation of Encryption Algorithms in MQTT Protocol for IoT Constrained devices
Iqbal, Y. W., Amjad, M. F., Khan, F. A., & Abbas, H. (2022) · International Conference on Computational Intelligence and Communication Networks (CICN) · doi:10.1109/cicn56167.2022.10008320 · Cited by 15
Survey of fault management techniques for edge-enabled distributed metaverse applications
Shaikh, S., & Jammal, M. (2024) · Computer Networks · doi:10.1016/j.comnet.2024.110803 · Cited by 14
Abuibaid, M., Ghorab, A. H., Seguin-Mcpeake, A., Yuen, O., Yungblut, T., & St-Hilaire, M. (2022) · IEEE Access · doi:10.1109/access.2022.3204976 · Cited by 11
Validation of High-Availability Model for Edge Devices and IIoT
Peniak, P., Bubeníková, E., & Kanáliková, A. (2023) · Sensors · doi:10.3390/s23104871 · Cited by 11
Neagu, M., Serban, C. M., Hângan, A., & Sebestyen, G. (2026) · Telecom · doi:10.3390/telecom7010010 · Cited by 11
Self-citationYakubov, D., & Hästbacka, D. (2025) · Lecture Notes in Computer Science · doi:10.1007/978-3-031-84617-5_8 · Cited by 10
CAROL: Confidence-Aware Resilience Model for Edge Federations
Tuli, S., Casale, G., & Jennings, N. R. (2022) · 2022 52nd Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN) · doi:10.1109/dsn53405.2022.00016 · Cited by 9
MOSE: A Novel Orchestration Framework for Stateful Microservice Migration at the Edge
Calagna, A., Yu, Y., Giaccone, P., et al. (2025) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2025.3579051 · Cited by 7
Akshatha, P. S., Hiremath, D., & Kumar, S. (2024) · doi:10.1109/icwite59797.2024.10503309 · Cited by 6
Function Offloading and Data Migration for Stateful Serverless Edge Computing
Nardelli, M., & Russo, G. R. (2024) · ACM/SPEC International Conference on Performance Engineering (ICPE) · doi:10.1145/3629526.3649293 · Cited by 5
Failover Timing Analysis in Orchestrating Container-based Critical Applications
Barletta, M., De Simone, L., Corte, R. D., & Di Martino, C. (2024) · 2024 19th European Dependable Computing Conference (EDCC) · doi:10.1109/edcc61798.2024.00026 · Cited by 5
Fault-Tolerant IoT System Using Software-Based “Digital Twin”
Baranwal, T., Varada, S., Das, S., & Haider, M. R. (2024) · 2024 IEEE 10th World Forum on Internet of Things (WF-IoT) · doi:10.1109/wf-iot62078.2024.10811168 · Cited by 5
Proactive Fault Tolerant Service Management in Edge Computing
Tambe, S. D., & A., A. F. (2023) · 2023 IEEE International Conference on Advanced Networks and Telecommunications Systems (ANTS) · doi:10.1109/ants59832.2023.10469661 · Cited by 2
Serban, C. M., Neagu, M., Hângan, A., et al. (2025) · International Conference on Control Systems and Computer Science · doi:10.1109/cscs66924.2025.00066 · Cited by 2
Self-citationPureEdge AI-Driven Scheduling for Failover Management in Edge Computing Systems
Dong-Seong, K., & Syamsul, R. (2025) · 2025 International Conference on Mobile, Military, Maritime IT Convergence (ICMIC) · doi:10.1109/icmic66299.2025.11257787 · Cited by 1
Li, T., Tian, L., Huang, G., & Li, A. (2025) · 2025 7th International Conference on Electronic Engineering and Informatics (EEI) · doi:10.1109/eei67650.2025.11334996 · Cited by 1
Analysis of Fault Management Techniques for Edge-Enabled Distributed Metaverse ApplicationsPreprint
Shaikh, S., & Jammal, M. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4882073
Research works printed but not linked: 1
- Du, 2020, Fault-Tolerating Edge Computing with Server Redundancy Based on a Variant of Group Degree Centrality, Proceedings of the International Conference on Service-Oriented Computing (ICSOC 2020); Lecture Notes in Computer Science, Volume 12571, 198
Cited by
Digital Twin Technology in Endoscopy: Current Applications, Challenges, and Future Perspectives
David-Olawade, A. C., Fidelis, S. C., Olawuyi, O. F., et al. (2026) · Gastroenterology & Endoscopy · doi:10.1016/j.gande.2026.06.005
Hannah, M., Ezeh, G. N., John, C., et al. (2026) · Advances in Networks · doi:10.11648/j.net.20261301.13
An Edge-First Digital Twin–Driven Cyber–Physical Home System for Resilient Aging-in-Place
Wang, Z., & Hu, X. (2026) · Sensing and Imaging · doi:10.1007/s11220-026-00851-2
“The selected studies cover wearable and ambient monitoring, robot-assisted home care, privacy-preserving home-care DTs, GenAI-enhanced DT interfaces, medical DT-CPS platforms, and edge-DT availability mechanisms [5 , 7 – 9 , 20 , 22 , 28 – 30].”
Metadata
- Authors
- Neagu, M., Serban, C. M., Hangan, A., & Sebestyen, G.
- Affiliations
- Department of Computer Science, Faculty of Automation and Computer Science, Technical University of Cluj-Napoca, 400027 Cluj-Napoca, Romania
- Technical University of Cluj-Napoca (RO)
- Venue
- Future Internet
- Publisher
- MDPI AG
- Published
- 2026-03-06
- Volume
- 18
- Issue
- 3
- Pages
- 137
- ISSN
- 1999-5903 (ISSN-L)
- OpenAlex
- W7134065245
- Cited by
- 3
- References
- 35
- Citations by year
- 2026: 3
- FWCI
- 7.11
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Future Internet (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- “Disponibilitate, confidențialitate și detecție de anomalii în sisteme IoT-DeCiD”
- Competition-GNaC ARUT 2023
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › Distributed systems and fault tolerance
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Engineering › Industrial and Manufacturing Engineering › Digital Transformation in Industry
- Keywords
- edge computing, IoT sensor networks, high availability, redundancy, state synchronization, fault tolerance, single point of failure, resource-constrained devices, fault injection
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-12, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-12 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.