Berkeley CNN finds 72 unseen pulses from FRB 121102

AI system

System: Convolutional neural network for FRB pulse detection (Zhang et al., Breakthrough Listen / Berkeley SETI Research Center)
Model: Not assessed
Release: Not assessed
Application developer: University of California, Berkeley (Berkeley SETI Research Center, Breakthrough Listen)
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: A long, high-time-resolution radio observation of a known repeater. A CNN trained on simulated and real pulses. Re-analysis of data already searched by classical methods.
Human oversight: Not assessed
Relation to AI: The CNN flagged 72 pulses in data that a classical de-dispersion search had already processed, finding 21. The detections were vetted by the authors and accepted in peer review. The result is a verified discovery, but the wording “independently confirmed” is not met. No re-detection by another group is among the sources.
AI contribution: The CNN flagged 72 pulses in data that a classical de-dispersion search had already processed, finding 21. The detections were vetted by the authors and accepted in peer review. The result is a verified discovery, but the wording “independently confirmed” is not met. No re-detection by another group is among the sources.

Outcome identifiers: Not assessed
Benefit: found 72 previously undetected pulses from the repeating fast radio burst FRB 121102
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

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: found 72 previously undetected pulses from the repeating fast radio burst FRB 121102
Comparison: the conventional search algorithm previously run on the same observation
Results: Reference: 21 bursts have been previously detected. Result: 72 new pulses, totaling 93 pulses in five hours.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Participants: Developer: University of California, Berkeley (Berkeley SETI Research Center, Breakthrough Listen). Operator: Breakthrough Listen, on Green Bank Telescope data.
Jurisdictions: Benefit: Berkeley, United States. Developer: United States. Operator: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Conditions for replication: Candidate pulses must survive radio-frequency-interference rejection and visual vetting. Independent confirmation would require re-detection by a second pipeline or group.
Replication attempts: Not assessed
Later findings: No later re-detection or follow-up of these 72 pulses 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

Assessed benefits: Organizations. Society.
Depth: Organizations: 2. Society: 2.
Reasons: Organizations: Breakthrough Listen recovered 72 additional bursts from an existing 400-terabyte Green Bank dataset without new telescope time. Society: The FRB research community gained a much larger burst sample (93 pulses in five hours) from the only known repeating FRB at the time.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: The deployer's account

Kind: Not assessed
Related entries: Not assessed
Field: Astrophysics / radio transients. Academic research.
Purpose: Built to detect fast radio burst pulses in Breakthrough Listen radio data that classical search algorithms miss. The 72 bursts were that purpose.
Dates: Date basis unassessed: 2018 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: 16a00c056b75…

Peer-reviewed study Developer

Zhang, Y. G., Gajjar, V., Foster, G., Siemion, A., Cordes, J., Law, C. and Wang, Y. (2018). Fast Radio Burst 121102 Pulse Detection and Periodicity: A Machine Learning Approach. The Astrophysical Journal 866, 149 (arXiv:1809.03043).
Press release Deployer

Green Bank Observatory (2018). AI Uncovers 72 New FRBs in GBT's Breakthrough Listen Data. Press release.