Hallucinations in Large Language Models (LLMs), defined as the generation of content inconsistent with facts or context, represent a core obstacle to their reliable deployment in critical domains. Current research primarily focuses on binary "detection" approaches that, while capable of identifying hallucinations, fail to provide interpretable and actionable feedback for model improvement, thus limiting practical utility. To address this limitation, a new research paradigm is proposed, shifting from "detection" to "diagnosis". The Hallucination Diagnosis Task is introduced, a task which requires models to not only detect hallucinations, but also perform error localization, causal explanation, and content correction. We develop the Hallucination Diagnosis Generator (HDG), an automated pipeline that systematically generates high-quality training samples with rich diagnostic metadata from raw corpora through multi-dimensional augmentation strategies including controlled fact fabrication and reasoning chain perturbation. Using HDG-generated data, we train HDM-4B-RL, a 4-billion-parameter hallucination diagnosis model, employing Group Relative Policy Optimization (GRPO) with a comprehensive reward function incorporating structural, accuracy, and localization signals. Experimental results demonstrate that our model surpasses previous state-of-the-art detection models on the HaluEval benchmark while achieving comparable performance to advanced general-purpose models. In comprehensive diagnosis tasks, HDM-4B-RL matches the capabilities of larger general models while maintaining a smaller size. This work validates the feasibility and value of hallucination diagnosis, providing an effective methodology for building more trustworthy and reliable generative AI systems.

LLM-based multi-agent systems have demonstrated significant capabilities across diverse domains. However, the task performance and efficiency are fundamentally constrained by their collaboration strategies. Prevailing approaches rely on static topologies and centralized global planning, a paradigm that limits their scalability and adaptability in open, decentralized networks. Effective collaboration planning in distributed systems using only local information thus remains a formidable challenge. To address this, we propose BiRouter, a novel dual-criteria routing method for Self-Organizing Multi-Agent Systems (SO-MAS). This method enables each agent to autonomously execute "next-hop" task routing at runtime, relying solely on local information. Its core decision-making mechanism is predicated on balancing two metrics: (1) the ImpScore, which evaluates a candidate agent's long-term importance to the overall goal, and (2) the GapScore, which assesses its contextual continuity for the current task state. Furthermore, we introduce a dynamically updated reputation mechanism to bolster system robustness in untrustworthy environments and have developed a large-scale, cross-domain dataset, comprising thousands of annotated task-routing paths, to enhance the model's generalization. Extensive experiments demonstrate that BiRouter achieves superior performance and token efficiency over existing baselines, while maintaining strong robustness and effectiveness in information-limited, decentralized, and untrustworthy settings.

Figure 1: EchoSight leverages in-situ optical backscatter tethering to achieve a look-and-control bidirectional interaction on commercial AR glasses.This process mimics human's natural interaction with low mental burden.EchoSight requires no pre-registration or network connection, facilitating scenarios where interactions are opportunistic and impromptu.

Transfer learning plays a crucial role in low-resource machine translation by addressing the challenge of poor model performance due to limited data in low-resource languages, thereby improving translation accuracy. Current research methods not only utilize pre-trained parent models for parameter initialization and fine-tuning but also use the soft labels output by these parent models to enhance the consistency between parent and child models. However, even if the parent model performs well, there are still instances where certain token predictions are unstable. During training, if the child model incorporates these unstable token predictions, it can hinder its learning effectiveness; the child model might not fully comprehend the parent model’s prediction strategy, potentially affecting overall translation performance. To address this, we propose a training strategy called Token-Level Filter Training, designed to effectively filter out unstable token predictions from the parent model, thereby transferring the parent model’s positive knowledge to the child model. Additionally, we introduce a hierarchical ranking loss method to help the child model better learn the parent model’s prediction strategies and sequence order, thus enhancing translation accuracy and fluency. Experimental results show that our method outperforms baseline methods on the public datasets Global Voices (Id, Ca, Hu, Pl) and WMT17 (Turkish–English), with BLEU score improvements of 1.47, 0.91, 0.50, 0.54, and 0.55, respectively. These results demonstrate the effectiveness and superiority of the proposed method.