Self-Knowledge Distillation (SKD) leverages the student's own knowledge to create a virtual teacher for distillation when the pre-trained bulky teacher is not available. Whilst existing SKD approaches demonstrate gorgeous efficiency in single-label learning, to directly apply them to multi-label learning would suffer from dramatic degradation due to the following inherent imbalance: \textit{targets with unified labels but multifarious visual scales are crammed into one image, resulting in biased learning of major targets and disequilibrium of precision-recall}. To address this issue, this paper proposes a novel SKD method for multi-label learning named Multi-label Self-knowledge Distillation (MSKD), incorporating three Spatial Decoupling mechanisms (i.e. Locality-SD (L-SD), Reconstruction-SD (R-SD), and Step-SD (S-SD)). L-SD exploits relational dark knowledge from regional outputs to amplify the model's perception of visual details. R-SD reconstructs global semantics by integrating regional outputs from local patches and leverages it to guide the model. S-SD aligns outputs of the same input at different steps, aiming to find a synthetical optimizing direction and avoid the overconfidence. In addition, MSKD combines our tailored loss named MBD for balanced distillation. Exhaustive experiments demonstrate that MSKD not only outperforms previous approaches but also effectively mitigates biased learning and equips the model with more robustness.

Large language models (LLMs) demonstrate outstanding capability in understanding natural language and show great potential in open-domain travel planning. However, when confronted with multi-constraint itineraries, personalized recommendations, and scenarios requiring rigorous external information validation, pure LLM-based approaches lack rigorous planning ability and fine-grained personalization. To address these gaps, we propose the Symbolic LoRA Travel Planner (SLTP) framework—an agent architecture that combines a two-stage symbol-rule LoRA fine-tuning pipeline with a user multi-option heuristic tree search (MHTS) planner. SLTP decomposes the entire process of transforming natural language into executable code into two specialized, sequential LoRA experts: the first maps natural-language queries to symbolic constraints with high fidelity; the second compiles symbolic constraints into executable Python planning code. After reflective verification, the generated code serves as constraints and heuristic rules for an MHTS planner that preserves diversified top-K candidate itineraries and uses pruning plus heuristic strategies to maintain search-time performance. To overcome the scarcity of high-quality intermediate symbolic data, we adopt a teacher–student distillation approach: a strong teacher model generates high-fidelity symbolic constraints and executable code, which we use as hard targets to distill knowledge into an 8B-parameter Qwen3-8B student model via two-stage LoRA. On the ChinaTravel benchmark, SLTP using an 8B student achieves performance comparable to or surpassing that of other methods built on DeepSeek-V3 or GPT-4o as a backbone.

Existing remote sensing object detection methods struggle with challenges such as complex background interference, variable object scales, and class imbalance due to a lack of coordinated internal optimization. This paper proposes AFDNet, a novel RSOD algorithm that establishes an internal collaborative evolution mechanism to systematically enhance the model’s feature perception and localization capabilities in complex scenes. AFDNet achieves this through three tightly coupled, co-evolving components: (1) A channel–spatial dual-sensing module that adaptively focuses on crucial features and suppresses background noise. (2) A dynamic bounding box optimization module that integrates distance-aware and scale-normalization strategies, significantly boosting localization accuracy and regression robustness for multi-scale objects. (3) A Gaussian adaptive activation unit that enhances the model’s nonlinear fitting capability for better detail extraction under weak conditions. Extensive experiments on two public datasets, RSOD and NWPU VHR-10, verify the excellent performance of AFDNet. AFDNet achieved a leading 95.16% mAP@50 on the RSOD dataset and an astonishing 96.52% mAP@50 on the NWPU VHR-10 dataset, which is significantly better than the mainstream detection models. This study verifies the effectiveness of introducing internal co-evolution mechanisms and provides a novel and reliable solution for high-precision remote sensing target detection.

With the rapid advancement of Neural Machine Translation (NMT), enhancing translation efficiency and quality has become a focal point of research. Despite the commendable performance of general models such as the Transformer in various aspects, they still fall short in processing long sentences and fully leveraging bidirectional contextual information. This paper introduces an improved model based on the Transformer, implementing an asynchronous and segmented bidirectional decoding strategy aimed at elevating translation efficiency and accuracy. Compared to traditional unidirectional translations from left-to-right or right-to-left, our method demonstrates heightened efficiency and improved translation quality, particularly in handling long sentences. Experimental results on the IWSLT2017 dataset confirm the effectiveness of our approach in accelerating translation and increasing accuracy, especially surpassing traditional unidirectional strategies in long sentence translation. Furthermore, this study analyzes the impact of sentence length on decoding outcomes and explores the model's performance in various scenarios. The findings of this research not only provide an effective encoding strategy for the NMT field but also pave new avenues and directions for future studies.

Maintaining high-fidelity character personas and tracking trusted narrative facts remain significant challenges for LLM-based role-playing systems, particularly in long-context scenarios. Traditional Retrieval-Augmented Generation (RAG) approaches, which typically rely on static, stateless retrieval, often struggle to capture evolving plot dynamics, leading to character hallucinations and logical inconsistencies over prolonged interactions. To address these limitations, we present FictionRAG, a novel stateful retrieval-augmented framework designed to enhance long-narrative role-playing. FictionRAG introduces a hierarchical memory architecture that decouples narrative information into three distinct lanes: factual events, persona traits, and worldview constraints. Furthermore, it employs a failure-driven metacognitive regulatory loop that dynamically identifies and corrects retrieval deficiencies—such as persona drift or conflicting world rules—before response generation. By treating role-playing as a dynamic state tracking problem rather than simple question answering, FictionRAG ensures that generated responses are strictly grounded in both the narrative timeline and the character’s psychological profile. Extensive experiments on a dataset comprising twenty classic novels demonstrate that FictionRAG significantly outperforms existing baselines in factual accuracy, persona stability, and worldview consistency. Beyond literary role-playing, these results suggest that stateful, evidence-constrained retrieval can serve as a general mechanism for long-form controllable generation tasks that require persistent state tracking and multi-dimensional consistency.