SNAI 2025 · 2nd Symposium on Networking and Artificial Intelligence · 24–26 October 2025, Singapore
This call closed on 27 June 2025. See the SNAI 2026 call for papers for the current edition.
SNAI 2025 invited original research contributions on the two-way intersection of computer networking and artificial intelligence. Papers were reviewed double-blind by at least three independent reviewers, with an author rebuttal phase and a dedicated artifact evaluation track (introduced at the 2025 edition).
Learned congestion control, traffic engineering, routing, scheduling, and resource allocation; anomaly and fault detection; intent-based and self-driving networks; closed-loop network automation; telemetry summarization; configuration synthesis; and operator-facing AI tools. This was the largest track in 2025, with 8 papers accepted.
Model placement and partitioning across the edge and the data plane; in-switch and SmartNIC inference; line-rate feature extraction; latency-aware serving; hardware–software co-design for in-network ML; and programmable pipelines for inference offload. This track accepted 7 papers.
Communication-efficient and privacy-preserving training; gradient compression and quantization; federated learning under heterogeneous and bandwidth-constrained links; straggler mitigation; decentralized training over peer-to-peer topologies; and split learning across edge and cloud. This track accepted 6 papers.
Intrusion and anomaly detection; encrypted-traffic classification; attack attribution; evasion and adversarial robustness; large-scale measurements of malicious traffic; and ML-assisted threat hunting. This track accepted 6 papers and produced the Best Student Paper award winner.
Programmable data planes (P4, eBPF, SmartNICs) and the algorithms that run on them; data-plane feature extraction; line-rate inference; match-action pipelines for ML; and benchmarks for data-plane ML workloads. This track was new in 2025 and accepted 6 papers.
Large language models for network operations, configuration synthesis, troubleshooting, and incident response; foundation models for telemetry and traffic analysis; agentic and tool-using models for network management; retrieval-augmented generation over operational knowledge bases; and evaluation methodology for generative tooling. This track was new in 2025, accepted 5 papers, and produced the inaugural Best Artifact award winner.
SNAI 2025 used the standard IEEE conference template, two-column, 10pt, US Letter or A4. LaTeX and Microsoft Word templates were available from the IEEE Conference Author Kit.
All papers were reviewed double-blind. Authors were required to anonymize their submission, including references phrased in the third person. Each paper received at least three independent reviews. Authors were able to respond to reviewers during the rebuttal period (8–15 August 2025). The program chairs resolved disagreements and consulted additional reviewers for borderline cases. Final decisions were issued on 22 August 2025.
The 2025 edition introduced artifact evaluation as an optional companion to paper submission. Authors of accepted papers could submit a separate artifact package — code, datasets, configuration files, or measurement traces — supporting the claims of their paper. Artifacts were evaluated by a separate artifact committee against criteria of availability, functionality, reproducibility, and documentation. The artifact review did not affect the paper's acceptance decision. Papers with successful artifact submissions were eligible for the inaugural Best Artifact award.
Accepted papers were published in the open-access SNAI 2025 proceedings, archived in the LiteConf repository, and assigned DOIs. Authors retain copyright. The proceedings were submitted to major indexing services.
For questions about the call, the program chairs were reachable at [email protected]. The full CFP record is preserved at liteconf.org/cfp/snai-2025/.