Spectrum deception has found broad utility across multiple domains, including electronic warfare, tactical counter-measures, and adversarial sensing suppression. However, generating complex time–frequency signatures relies on sophisticated signal processing pipelines, which poses significant challenges for UAV and other IoT platforms with severely constrained onboard computational resources. Moreover, the limited onboard capability further restricts rapid signal synthesis and adaptation, failing to meet the strict rapid-response requirements of the battlefield. To address this dilemma, we propose the Fast-Tactical Signal Deception Framework (FT-SDF), a specialized generative architecture optimized for real-time signal synthesis. We formulate a novel spectro-temporal diffusion dynamics mechanism that innovatively incorporates additional spectral blurring and reverse process variance, jointly optimizing noise prediction and variance, which is necessary to preserve fine-grained spectral structures and key time–frequency signatures across different modulation schemes. Notably, to ensure strict adherence to communication protocols, we introduce a lightweight spectrum-context encoder that employs a dual-domain embedding strategy for physics-aware conditioning. Furthermore, to enable rapid inference, we develop a variance-aware acceleration mechanism that exploits learned spectral uncertainty to guide a dynamic warm-start schedule, thereby drastically compressing the sampling trajectory. Extensive evaluations on a systematically reconstructed RadioML benchmark demonstrate that FT-SDF outperforms state-of-the-art baselines. Specifically, it achieves a 59.3% reduction in sampling iterations (more than 2-fold inference speedup), while maintaining both high statistical fidelity (FID < 15) and industrial-grade precision (EVM ≤ 14%), demonstrating the feasibility of rapid, controllable generative AI in complex electromagnetic environments.

Fine-tuning pretrained language models on helpful intermediate tasks often greatly improves the performance of target tasks.However, how to efficiently find the source tasks that can successfully transfer still remains underexplored.In this work, we propose to learn an affinity scoring function to predict transferability between tasks.Specifically, we conduct prompt tuning and regard soft prompts as task embeddings that summarize task-specific information.Then we randomly sample task pairs to train an affinity scoring function.The goal is to predict the transfer gain (i.e., affinity) between a task pair, by conditioning on their task embeddings.Once the scoring function is trained, given a novel target task, we use it to predict the most transferable source tasks, without a brute-force search for all possible source-target pairs.Experimental results across 50 tasks show that our method efficiently identifies beneficial tasks for transfer learning.

Despite pre-trained language models such as BERT have achieved appealing performance in a wide range of Natural Language Processing (NLP) tasks, they are computationally expensive to be deployed in real-time applications. A typical method is to adopt knowledge distillation to compress these large pre-trained models (teacher models) to small student models. However, for a target domain with scarce training data, the teacher can hardly pass useful knowledge to the student, which yields performance degradation for the student models. To tackle this problem, we propose a method to learn to augment data for BERT Knowledge Distillation in target domains with scarce labeled data, by learning a cross-domain manipulation scheme that automatically augments the target domain with the help of resource-rich source domains. Specifically, the proposed method generates samples acquired from a stationary distribution near the target data and adopts a reinforced controller to automatically refine the augmentation strategy according to the performance of the student. Extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art baselines on different NLP tasks, and for the data-scarce domains, the compressed student models even perform better than the original large teacher model, with much fewer parameters (only ~13.3%) when only a few labeled examples available.

Transfer learning (TL) seeks to improve the learning of a data-scarce target domain by using information from source domains.However, the source and target domains usually have different data distributions, which may lead to negative transfer.To alleviate this issue, we propose a Wasserstein Selective Transfer Learning (WSTL) method.Specifically, the proposed method considers a reinforced selector to select helpful data for transfer learning.We further use a Wasserstein-based discriminator to maximize the empirical distance between the selected source data and target data.The TL module is then trained to minimize the estimated Wasserstein distance in an adversarial manner and provides domain invariant features for the reinforced selector.We adopt an evaluation metric based on the performance of the TL module as delayed reward and a Wasserstein-based metric as immediate rewards to guide the reinforced selector learning.Compared with the competing TL approaches, the proposed method selects data samples that are closer to the target domain.It also provides better state features and reward signals that lead to better performance with faster convergence.Extensive experiments on three real-world text mining tasks demonstrate the effectiveness of the proposed method.