Federated learning (FL) enables collaborative training across distributed clients without directly sharing their raw data. However, non-IID data distributions and malicious behaviors can produce unreliable and conflicting client updates, severely degrading aggregation stability and global generalization. Existing robust aggregation methods often depend on additional trust assumptions, server-side clean data, or expensive pairwise update comparisons, which not only limit their practicality but also increase computational and resource overhead in real-world FL scenarios. To address these limitations, we propose FedRGEA, a reliability-guided bio-inspired evolutionary aggregation framework for resource-aware robust federated aggregation. FedRGEA integrates self-calibrated reliability assessment (SCRA) with reference-guided conflict-aware evolutionary aggregation (RCEA). Specifically, SCRA constructs dynamic references from client updates themselves and evaluates update reliability through multi-criteria deviation analysis without relying on externally trusted references. The resulting reliability priors are incorporated into RCEA, which constructs a refined reference from the retained updates and formulates robust aggregation as a simplex-constrained evolutionary weight search problem by jointly considering reliability priors, reference-based conflicts, and weight concentration. Experimental results show that FedRGEA outperforms representative FL and robust aggregation baselines under the 40% sign-flipping attack, attaining accuracies of 77.88%, 83.97%, and 97.92% on ISIC 2019, HAM10000, and BrainTumor, respectively. These results demonstrate the effectiveness of FedRGEA in maintaining robust aggregation under a high proportion of malicious clients.

Multi-label text feature selection aims to identify compact and discriminative feature subsets for documents associated with multiple correlated labels. In federated environments, this task is further complicated by high-dimensional search spaces, non-independent and non-identically distributed data, complex label dependencies, and the requirement that raw feature and label data remain local. To address these challenges, this paper proposes the Federated Anchor-Guided Frequency-Domain Optimization Algorithm (Fed-AGFDO), a federated multi-label text feature selection framework. At each client, the Anchor-Guided Frequency-Domain Optimization (AGFDO) module transforms the encoded search population into the frequency domain and combines random spectral exploration, elite-guided mixing, diversity-aware anchor screening, and multi-anchor direction aggregation to balance exploration and exploitation. Candidate feature-weight vectors are evaluated using a manifold-regularized fitness function that jointly considers sample structure, label-consensus regularization, sparsity regularization, and a soft global feature-weight consistency penalty. The server aggregates only local feature-weight vectors through sample-size-weighted aggregation followed by an exponential moving average update to smooth inter-round changes in the global feature weights while keeping raw data local. Experiments on eight multi-label text datasets demonstrate that Fed-AGFDO achieves strong classification and label-ranking performance with compact feature subsets. Under the representative subset settings, Fed-AGFDO improves Average Precision by up to $5.83\%$ over Fuzzy Federated Multi-Label Feature Selection (Fuzzy FMFS) and reduces Ranking Loss by up to $22.65\%$ relative to Federated Multi-Label Feature Selection (FMLFS). Parameter sensitivity and ablation analyses further demonstrate robustness across the examined parameter ranges and the complementary contributions of anchor guidance, diversity screening, and multi-anchor aggregation. These results indicate that Fed-AGFDO provides an effective and data-locality-aware solution for high-dimensional federated multi-label text feature selection.