Kernel methods, which embed data distributions into a reproducing kernel hilbert space (RKHS) via positive-definite similarity measures, continue to play an important role. However, learning a good, generalizable kernel for high-dimensional and heterogeneous data under temporal or regional distribution shift remains challenging. To address these issues, we propose SpectraMancer, which learns kernels directly in the Fourier spectral domain induced by multilevel circulant matrices, thereby enabling generalizable kernel learning for complex data. SpectraMancer embeds all shift-invariant candidates into a common multilevel order via randomized multilevel circulant matrices, which yields a fixed Fourier diagonalization and turns inverses, products, and gradients into elementwise fast Fourier transform (FFT) operations. To the best of our knowledge, this is the first kernel-learning approach that exploits randomized multilevel circulant matrices for joint diagonalization across kernels. SpectraMancer further enforces scale invariance via kernel double centering and Frobenius normalization, reduces spectral variance through antithetic phase pairing with quasi-Monte Carlo draws, and optimizes a solver-free spectral risk proxy (SRP) for bandwidth weighting without repeated inner solves. Experimental results show that SpectraMancer improves spectrum-aware kernel selection and predictive performance across diverse benchmarks.
Lizhong Ding, Jiarun Fu, Qiuning Wei et al.· IEEE Transactions on Neural...· 0 citations
Large language model agents have shown strong potential in complex interactive tasks, yet their reinforcement learning (RL) is often hindered by sparse rewards, as a long multi-turn trajectory may receive only a single outcome-level signal. On-policy self-distillation (OPSD) provides dense token-level supervision from a privileged teacher, but the teacher may not be reliable at every position. Existing methods commonly rely on isolated token-level discrepancies, which can be sensitive to noise, or assign a shared step-level weight that may overlook positional variation. We propose Persistent Consistency Self-Distillation (PCSD), which derives token-level distillation weights from the local persistence of teacher-favoring signals. PCSD combines adaptive windows with exponentially decayed aggregation to capture persistent relative teacher support, applies trend-aware modulation to attenuate locally declining support, and produces continuous weights through sigmoid gating. The resulting objective is jointly optimized with GRPO, combining dense teacher guidance with sparse environmental feedback. Without inference-time skills, PCSD achieves the best ALFWorld Overall results among all baselines on both backbones, exceeding GRPO by 15.6 and 13.3 points and SDAR by 6.2 and 5.5 points, while remaining competitive on WebShop and gaining 15.8 points over GRPO on unseen ALFWorld split.
Chunji Lv, Yangguang Wei, Junlin Liu et al.· 0 citations
FedGAMMA is proposed, casting federated multimodal graph foundation learning as a two-stage semantic-structural alignment problem of federated pre-training and prompt-based fine-tuning, and outperforms competitive baselines accross multi-domain datasets on multiple tasks.
Xunkai Li, Guohao Fu, Yuming Ai et al.· 0 citations
Agent harnesses have become the operational infrastructure of modern large language model agents, coordinating context, tools, verification, and execution control to translate latent model capability into reliable long-horizon behavior. However, reliable long-horizon behavior requires harness control to adapt to task demands, execution environments, and evolving execution states, whereas current harnesses predominantly rely on hand-crafted or globally fixed policies; this mismatch manifests as unnecessary computational overhead and, in adverse cases, reduced task success. To address this limitation, we formulate the task of enabling adaptive orchestration in harness systems as a causal learning problem and propose Counterfactual Harness Intervention Learning for Long-Horizon Agents (CHILL-Harness). CHILL-Harness intervenes at the orchestration layer to enable advantage-guided workflow adaptation, thereby improving reasoning and execution efficiency while preserving task performance. Specifically, we develop causal intervention effect learning as the effect-estimation component of CHILL-Harness to estimate intervention-relative workflow advantage from confidence-weighted execution evidence and identify advantageous workflow adaptations. We further introduce advantage-realizing causal orchestration as its realization component to adaptively allocate counterfactual reasoning and realize only workflow adjustments supported by sufficient expected advantage. Finally, we incorporate a success-preserving objective and advantage-margin authorization constraints into CHILL-Harness to promote reliable adaptation. Extensive experiments on heterogeneous long-horizon tasks spanning information seeking, software engineering, and terminal interaction show that CHILL-Harness consistently preserves or improves task success while substantially reducing token consumption and execution time.
Jiarun Fu, Lizhong Ding, Sida Chen et al.· 0 citations
