AresGW finds eight new gravitational-wave candidates in LIGO O3

AI system

System: AresGW (ResNet-based deep learning gravitational-wave detection pipeline)
Model: AresGW.
Release: Exact release not established.
Application developer: Aristotle University of Thessaloniki (Departments of Physics and Informatics) with the Swiss Data Science Center, ETH Zurich
Role: The pipeline detected and ranked possible gravitational-wave events.
Configuration: AresGW used a hierarchy of classifiers to assess possible signals. The 2025 follow-up calls the model AresGW model 1.
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: A ResNet trained on LIGO Hanford and Livingston data with hierarchical classification of triggers by noise and frequency filters. 6.7 months of coincident O3 data; significance from a logarithmic ranking statistic and injections into O3 data.
Human oversight: Not assessed
Relation to AI: The deep-learning pipeline produced the candidate list and the ranking statistic. The authors computed p_astro from injections into O3 data. The developers are the authors, and the only source for the discovery is their own peer-reviewed paper and preprint. The paper itself calls the events candidates and reports no external confirmation.
AI contribution: The deep-learning pipeline produced the candidate list and the ranking statistic. The authors computed p_astro from injections into O3 data. The developers are the authors, and the only source for the discovery is their own peer-reviewed paper and preprint. The paper itself calls the events candidates and reports no external confirmation.

Benefits

Outcome identifiers: Not assessed
Benefit: eight new O3 candidates with p_astro > 0.5 (cumulative 5.94)
Category: Improved Knowledge Discovery
Status: Not assessed
Beneficiary groups: Organizations. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: eight new O3 candidates with p_astro > 0.5 (cumulative 5.94)
Comparison: the matched-filtering catalogs (all GWTC pipelines, OGC, IASa and IASb) within AresGW's effective training range
Results: Reference: 36 events (all GWTC pipelines); 37 (OGC); 34 (IASa); 36 (IASb). Result: 42.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: Aristotle University of Thessaloniki (Departments of Physics and Informatics) with the Swiss Data Science Center, ETH Zurich. Operator: The same research group, on public LIGO O3 data.
Jurisdictions: Benefit: Thessaloniki, Greece. Developer: Greece. Operator: Greece.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Sources within the effective training range (component masses 7 to 50 solar masses, chirp masses 10 to 40 solar masses). Two-detector coincidence; the network was not trained on Virgo data. A follow-up by the same group found sensitivity varies notably between month-long datasets, so detection counts may not transfer.
Replication attempts: 1 follow-up entries. Outcome unresolved: 1.
Later findings: The group's September 2025 follow-up re-evaluates the earlier 'AresGW model 1', not the model 2 that produced the eight candidates, and no confirmation or refutation of the candidates was found in the cited sources.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0282-source-0001–0003
Citations: Numbered references below.
Language: Not assessed
Source relationships: Developer
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: Organizations. Society.
Depth: Organizations: 2. Society: Not assessed.
Reasons: Organizations: The group's pipeline found eight new O3 candidates with p_astro above 0.5 beyond the catalogs it compares against, a measurable gain in detection capability. None is confirmed by another pipeline. Society: If real, the candidates add to the binary black hole population; no catalog has adopted them in the sources.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: The makers' claim

Entry details

Kind: Not assessed
Related entries: Not assessed
Field: Astrophysics / gravitational-wave astronomy. Academic research on public detector data.
Purpose: AresGW was built to detect binary black hole signals in LIGO data faster than matched filtering. Finding new candidates was that purpose.
Dates: Date basis unassessed: 2025 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: 44a8bcc77c34…

References

Other identifiers: Not assessed
Category mappings: Not assessed
Brief: Not assessed
Revisions: 7 revisions. Latest: 21 October 2026.

Sources

Peer-reviewed study Developer

Koloniari AE, Koursoumpa EC, Nousi P, Lampropoulos P, Passalis N, Tefas A, Stergioulas N (2025). New gravitational wave discoveries enabled by machine learning. Machine Learning: Science and Technology 6(1):015054. Open access, CC BY 4.0. (Authored by the pipeline's developers.)
Preprint Developer

Koloniari AE et al. (2024). New Gravitational Wave Discoveries Enabled by Machine Learning. arXiv:2407.07820, 10 July 2024 (table of the eight candidates with p_astro and false alarm rates). (Authored by the pipeline's developers.)
Preprint Developer

Koloniari AE, Lazaridis L, Paschalidis C, Stergioulas N (2025). Robustness of Sensitivity Evaluations for Gravitational Wave Detection Algorithms. arXiv:2509.05283, 5 September 2025. (Authored by the pipeline's developers.)