Restarting Neutrino+Gamma Subgroup Meetings
Join the Zoom room
https://uv-es.zoom.us/j/69754566212?pwd=aZVSKJzZfERKG3FPFvpL8udWRNMapt.1
Join and take notes
https://uv-es.zoom.us/j/69754566212?pwd=aZVSKJzZfERKG3FPFvpL8udWRNMapt.1&mynotes=on
Link to the chat
https://uv-es.zoom.us/launch/jc/69754566212
Meeting ID: 697 5456 6212
Password: 945820
Key Outcomes
The group reviewed outputs from the EuCAIF Heidelberg workshop and the Lorentz Center meeting to shape a roadmap for the Gamma-Ray/Neutrino working group. Two near-term meeting themes were agreed upon: (1) event reconstruction for gamma-rays and neutrinos and tranferable learning, and (2) multi-messenger source analysis. Coordination with the Multi-Messenger and Cutting-Edge Methods subgroups was flagged as essential to avoid duplication, particularly around foundation model development.
Decisions Made
• Next two meetings structured around: (1) event reconstruction (IceCube paper + Thomas Vuillaume's EuCAIF talk) and (2) source analysis (Youyou Li's SBI work + gamma-ray worsk with Fermi-LAT and CTAO )
• Invite Gamma-ray working team from the Lorentz Center Workshop
• Gamma-ray and neutrino tracks could run in parallel rather than sequentially, given complementary scope
• Coordinate with Cutting-Edge Methods group, which plans emulators for neutrino physics (e.g., replacing CORSIKA with diffusion models)
Key Technical Discussion
• EuCAIF breakdown: ~95% of talks were LHC-focused (jet tagging, event reconstruction); cosmology ~4%; astrophysics ~1%
• Neutrino challenges: low count statistics (e.g., NGC 1068 significance ~4σ, well below 5σ threshold), making source localization difficult
• Youyou Li's SBI approach: applied CNN + SBI to a toy IceCube-like model with diffuse + unresolved source components; injected values recovered with reasonable posterior intervals
• Three differentiated pipeline steps identified: (1) event reconstruction, (2) source identification and characterization, (3) dedicated source emulator
• Foundation model scope: likely requires at least three separate models for the three pipeline stages, not one unified model
Cross-Group Alignment
• Multi-Messenger group (Nachiketa + Peter + Xiaofei): building simulated kilonova dataset combining gravitational wave and EM signatures for source classification — bottom-up approach 11
• Gravitational waves subgroup: ~10 active members, bi-weekly meetings, pipeline and foundation model strategy already defined
• Shared interest in latent space representation methods across both working groups
• Judit and Suvo to will join multi-messenger group meeting
Pending Confirmation
• Overlap with Cutting-Edge Methods group on neutrino emulators needs explicit coordination to avoid redundancy
Open Questions
• Whether a foundation model trained across multiple experiments (gamma-ray + neutrino) is feasible given differing data formats and statistics
• How to combine gamma-ray and neutrino data to improve source localization (e.g., reducing search radius from ~5° to ~1°)
• Leptonic vs. hadronic production discrimination requires dedicated source physics simulation — manpower and scope not yet defined
Action Items
• Judit + Shuvo: Contact gamma-ray's Lorentz Workshop Team & Thomas Vuillaume; set up a Doodle to schedule their participation.
• All: Consider joining the next cross-cutting methods group meeting
• Nachiketa: Circulate update to broader mailing list once kilonova pilot dataset and initial tasks are finalized