Continual configuration tuning in cellular radio access networks (RANs) is critical for maintaining performance, reliability, energy efficiency, and user experience. However, this task remains largely manual in practice. Automating it needs to confront high-dimensional configuration spaces, sparse and biased exploration, strong parameter interactions, and substantial environmental confounding. Existing RAN configuration tuning approaches have limited effectiveness in addressing these challenges. In this paper, we present CausalTune, a novel causal learning framework for automated RAN configuration optimization based on observational telemetry. CausalTune disentangles configuration effects from environmental and operational confounders, generalizes to sparse and previously unseen parameter settings, and captures high-impact multi-parameter interactions. Our key insight is that effective causal inference in operational RANs requires reshaping raw telemetry to expose confounding and learning environment-invariant mechanisms. Guided by this insight, CausalTune employs a multistage pipeline that integrates distributional representation learning, causal modeling, and interaction-aware recommendation. We evaluate CausalTune using 10 months of RAN measurement data from 1M commercial cells of a major cellular operator. Our comparison to state-of-the-art baselines shows that CausalTune achieves up to 3X KPI improvement on the held-out dataset. In terms of causal modeling quality, CausalTune achieves up to 12X lower KPI reconstruction error; on recommended configuration safety, it achieves 4X higher agreement with expert engineers while significantly reducing off-target recommendations. These findings demonstrate the potential of causal learning to enable reliable, scalable, and interpretable RAN configuration tuning.

Foundation model training is increasingly centralized in large cloud data centers because it demands immense compute and memory resources. Training over decentralized edge devices could democratize this ecosystem by harnessing otherwise idle compute, but prior edge-training systems fall short: they scale poorly with model size and device count, exceed per-device memory budgets, incur prohibitive collective communication, and are fragile to heterogeneous and dynamic device availability. We present Cleave, a parameter-server-centric framework that makes tensor-parallel training practical at the edge. Cleave introduces selective hybrid tensor parallelism, which finely shards GEMM-dominated training operations into memory-feasible sub-tasks while avoiding peer-to-peer collectives that become bottlenecks on asymmetric edge links. A cost model guides device selection and shard placement to mitigate stragglers and rapidly adapt to churn. Across OPT and Llama2 models, Cleave matches cloud GPU training efficiency while scaling to thousands of devices. It supports up to 8× more devices than prior edge approaches, reduces per-batch training time by up to 10×, and achieves 100× faster recovery from device failures.

The sparse Mixture-of-Experts (MoE) architecture is increasingly favored for scaling Large Language Models (LLMs) efficiently, but it depends on heterogeneous compute and memory resources. These factors jointly affect system Cost, Accuracy, and Performance (CAP), making trade-offs inevitable. Existing benchmarks often fail to capture these trade-offs accurately, complicating practical deployment decisions. To address this, we introduce MoE-CAP, a benchmark specifically designed for MoE systems. Our analysis reveals that achieving an optimal balance across CAP is difficult with current hardware; MoE systems typically optimize two of the three dimensions at the expense of the third—a dynamic we term the MoE-CAP trade-off. To visualize this, we propose the CAP Radar Diagram. We further introduce sparsity-aware performance metrics—Sparse Memory Bandwidth Utilization (S-MBU) and Sparse Model FLOPS Utilization (S-MFU)—to enable accurate performance benchmarking of MoE systems across diverse hardware platforms and deployment scenarios. This benchmark is available on Github: https://github.com/sparse-generative-ai/MoE-CAP.

Understanding link-level performance is key to assuring the quality of cloud-based and OTT services, optimal path selection, robust network operations and beyond. However, direct measurement of each link not only incurs high overhead at the Internet-scale but also is infeasible due to lack of access to network measurement information beyond AS boundaries and functional limitations at relay nodes. Although network tomography is well suited, existing approaches are insufficient due to their unrealistic assumptions with respect to stability, controllability, and visibility. Motivated by this, we propose PAINT, an online iterative algorithm that estimates and refines link-level performance metrics based on path-level measurement. In PAINT, the link metrics are iteratively estimated by minimizing their least square error (LSE) and calibrated based on the comparison of weight between the estimated shortest paths (SPs) and best-known paths from end-to-end path measurements. The key insight is that when there is inconsistency between these paths, then weights of links on the estimated SP are likely mis-estimated, triggering a further round of estimation to refine the estimated link metrics. Evaluation of PAINT, focusing on link delay estimation, using four different real network topologies and two real-world measurement datasets (including one we collected) shows that relative to existing approaches, it yields up to 3x gain in absolute link delay estimation accuracy and improves decisions dependent on link delay estimation by up to 5x in relative error.