To overcome the obstacles of maintaining covert transmissions in wireless networks employing collaborative wardens, we develop a reinforcement learning framework that jointly optimizes cooperative jamming strategies and relay selection mechanisms. The study focuses on a multi-relay-assisted two-hop network, where potential relays dynamically act as information relays or cooperative jammers to enhance covertness. A reinforcement learning-based relay selection scheme (RLRS) is employed to dynamically select optimal relays for signal forwarding and jamming; the framework simultaneously maximizes covert throughput and guarantees warden detection failure probability, subject to rigorous power budgets. Numerical simulations reveal that the developed reinforcement learning approach outperforms conventional random relay selection (RRS) across multiple performance metrics, achieving (i) higher peak covert transmission rates, (ii) lower outage probabilities, and (iii) superior adaptability to dynamic network parameters including relay density, power allocation variations, and additive white Gaussian noise (AWGN) fluctuations. These findings validate the effectiveness of reinforcement learning in optimizing relay and jammer selection for secure covert communications under colluding warden scenarios.

Large language model (LLM) agents often need historical conversations to recover argument values omitted from current tool requests. This study introduces SCAR–Pair, a two-stage ranker for memory retrieval after tool selection. It separates schema-expanded candidate generation from parameter-level scoring and integrates request, tool, field, and cross-encoder evidence through query-balanced pairwise learning. On a reconstructed Mem2ActBench bank containing 19,661 conversations and 391 evidence-bearing tasks, SCAR–Pair achieved 0.450 Recall@5 and 0.365 nDCG@5 under stratified five-fold evaluation, compared with 0.396 and 0.325 for a matched request-only pairwise ranker. Component-aware intervals and adjusted tests support these task-weighted gains in the ToolACE-dominated mixture. Controlled comparisons support the complete schema-view scoring configuration over request-only scoring on the same candidate pool but do not isolate correct-schema semantics or individual field components as independent contributors. The pairwise advantage over pointwise training is also not significant. Across the tested readers, higher retrieval quality does not yield a significant tool-call accuracy advantage over the strict request-only control. SCAR–Pair therefore improves selected-tool memory evidence ranking within the evaluated setting.