Qingyi Guo | Search arXiv

arXiv2026

Per-Channel Launch-Power Optimization in Hollow-Core Fiber Systems

In single-mode fiber (SMF) the Kerr effect ties every channel's quality of transmission to its neighbors' launch powers through cross-phase modulation (XPM), four-wave mixing (FWM) and inter-channel stimulated Raman scattering (ISRS), forcing a jointly planned launch profile. Hollow-core fiber (HCF), with a Kerr coefficient three to four orders of magnitude below silica, turns the per-channel powers into nearly independent knobs limited only by the shared amplifier budget. We build a per-channel generalized signal-to-noise ratio (GSNR) budget for amplified HCF links, including amplified spontaneous emission (ASE) with a wavelength- and output-power-dependent erbium-doped fiber amplifier (EDFA) noise figure (NF), inter-modal interference (IMI), nonlinearity in amplifier pigtails, CO2 gas-line loss and a flat transceiver (TRx) noise ceiling, and derive a sensitivity law that predicts when power shaping pays: its gain is bounded by the ASE noise share, canceled by self-phase modulation (SPM) at the SMF single-channel optimum, and positive in HCF. Across 80x64-GBaud C-band links over 400-3200 km, per-channel optimization buys up to 1.0 dB of worst-channel GSNR over a flat launch as the EDFA NF spread grows to 4 dB, cuts cross-channel power sensitivity by more than two orders of magnitude relative to SMF, and reaches a given GSNR at about 3 dB lower aggregate amplifier output. At a fixed amplifier budget this becomes a 1.26 to 1.41 times worst-channel reach extension, against at most 15% on the nonlinearity-capped SMF link.

physics.optics↗

arXiv2026

OpsMem: Dual-Memory Reasoning with Cross-Memory Resonance for Failure Diagnosis

Failure diagnosis in modern software systems requires iterative evidence acquisition and hypothesis reasoning guided by operational experience. Existing LLM-based methods improve diagnosis through agentic reasoning or knowledge augmentation, but they often lack a mechanism to coordinate the evolving diagnostic state with operational experience during iterative diagnosis. We propose OpsMem, a dual-memory framework that maintains a short-term memory for the current diagnostic state and a long-term memory for reusable operational experience. OpsMem uses cross-memory resonance to activate state-relevant long-term memory, conditions multi-agent diagnosis on the short-term and activated long-term memories, and consolidates reusable experience from solved incidents back into long-term memory. Experiments on a real-world Huawei microservice failure diagnosis dataset show that OpsMem outperforms representative agentic-reasoning and knowledge-augmented baselines, improving Match and Relevant by up to 46.88% and 18.39% over the strongest baseline, respectively.

cs.AI↗