Explore the vast range of titles available.

Objectives Follicular thyroid cancer (FTC) is prone to distant metastasis, and patients with distant metastasis often have poor prognosis. In this study, the impact of metastasis and other relevant factors on the prognosis of follicular thyroid carcinoma was examined. Methods This was a retrospective study. Data were obtained from Zhejiang Cancer Hospital, Sun Yat‐sen University Cancer Center and Hangzhou First People's Hospital affiliated with Zhejiang University School of Medicine, from January 2009 to June 2021 for 153 FTC patients. The patients were assigned into three groups according to their distant metastasis: distant metastasis at initial diagnosis (M1), distant metastasis during follow‐up (M2), and no evidence of distant metastasis over the course of the study (M0). Data were collected and summarized on clinical data, laboratory parameters, imaging features, postoperative pathologic subtypes, and metastases. The Cox proportional hazard model was used to perform the univariate and multivariate analysis. Kaplan–Meier curves were used to evaluate cancer‐specific survival (CSS). Results Based on metastasis, the patients were assigned into three groups, including 31 in the M1 group, 15 in the M2 group, and 107 in the M0 group. These individuals were followed up for an average of 5.9 years, and the group included 46 patients with distant metastasis (31 confirmed at diagnosis and 15 found during follow‐up). Univariate Cox regression analysis showed that age, Hashimoto's thyroiditis (HT), surgery method, postoperative adjuvant therapy, histologic subtype, nodule size, calcification, TSH, and distant metastasis all impacted prognosis. Multivariate Cox regression analysis suggested that histologic subtype (widely invasive; HR: 7.440; 95% CI: 3.083, 17.954; p < 0.001), nodule size (≥40 mm; HR: 8.622; 95% CI: 3.181, 23.369; p < 0.001) and distant metastasis (positive; HR: 6.727; 95% CI: 2.488, 18.186; p < 0.001) were independent risk factors affecting the prognosis of follicular thyroid cancer. Conclusions Histologic subtype, nodule size, and distant metastasis are important risk factors for the prognosis of follicular thyroid cancer. Patients with metastatic follicular thyroid cancer have a poor prognosis, especially with metastasis at the time of initial diagnosis. As a result, this group of patients requires individualized treatment and closer follow‐up.

Purpose Contrast-enhanced ultrasound (CEUS) can distinguish between central lung cancer and atelectatic lung tissue. The aim of this study was to explore the clinical value of CEUS for biopsy in patients with central lung cancer with obstructive atelectasis. Methods One hundred and twelve patients were selected and CEUS was performed to display central lung cancer and atelectatic lung tissue. The front edge of central lung cancer was punctured with a needle, avoiding the necrotic area, under the guidance of CEUS. Results All of the 112 lesions were diagnosed with a clear central lung cancer mass and atelectatic lung tissue. In 104 cases, the central lung cancer mass presented with a “slow-in and fast-out” pattern compared to atelectatic lung tissue. In eight cases, the central lung cancer mass presented with a “fast-in and fast-out” pattern compared to atelectatic lung tissue. The mean number of punctures was 2.6, and the success rate of puncture biopsy was 98%. Of the 112 patients, six cases had hemoptysis during the procedure and 10 patients had bloody sputum in the postoperative period. No complications were found in the other cases. Conclusion CEUS has important clinical value for needle biopsy of central lung cancer with atelectasis.

Papillary thyroid cancer (PTC) accounts for more than 80% of thyroid cancers, and ultrasound (US) imaging is the preferred method for the diagnosis of PTC. However, accurate prediction of different patterns of cervical lymph node metastasis (CLNM) in PTC continues to be a challenge. Herein, US images and clinical factors of PTC patients from three hospitals for more than 11 years are collected, and a multimodal deep learning model called DeepThy‐Net is then developed to predict different CLNM patterns. The proposed model not only uses the convolutional features extracted by deep learning but also integrates traditional clinical factors that are highly related to lymph node metastasis. Finally, the model is tested in two independent test sets, and the experimental results show that the area under curve (AUC) is between 0.870 and 0.905, indicating clinical applicability. The proposed method provides an important reference for the treatment and management of PTC. Moreover, for PTC cases involving an active surveillance strategy, the proposed method can serve as an important CLNM early warning tool. A DeepThy‐Net model is built to extract the features of the ultrasound images and predict different cervical lymph node metastasis patterns in papillary thyroid cancer. The clinical factors recorded by doctors are also digitized and input into a fully connected network with the above‐mentioned features, and finally, the prediction results are obtained.

Although using artificial intelligence (AI) to analyze ultrasound images is a promising approach to assessing thyroid nodule risks, traditional AI models lack transparency and interpretability. We developed a multimodal generative pre-trained transformer for thyroid nodules (ThyGPT), aiming to provide a transparent and interpretable AI copilot model for thyroid nodule risk assessment and management. Ultrasound data from 59,406 patients across nine hospitals were retrospectively collected to train and test the model. After training, ThyGPT was found to assist in reducing biopsy rates by more than 40% without increasing missed diagnoses. In addition, it detects errors in ultrasound reports 1,610 times faster than humans. With the assistance of ThyGPT, the area under the curve for radiologists in assessing thyroid nodule risks improved from 0.805 to 0.908 ( p  < 0.001). As an AI-generated content-enhanced computer-aided diagnosis (AIGC-CAD) model, ThyGPT has the potential to revolutionize how radiologists use such tools.

Background To investigate the diagnostic value of ultrasound gray scale ratio (UGSR) in differentiating papillary thyroid microcarcinomas (PTMCs) from benign micronodules (BMNs) in patients with Hashimoto’s thyroiditis (HT). Methods The ultrasound images of 285 PTMCs (from 247 patients) and 173 BMNs (from 140 patients) in the HT group, as well as 461 PTMCs (from 417 patients) and 234 BMNs (from 197 patients) in the non-HT group were retrospectively analyzed. The diagnosis of all cases was confirmed by histopathological examinations. The gray scale values of the nodules and surrounding thyroid tissues were measured and subsequently the UGSRs were calculated. Receiver operating characteristic curve analysis was used to determine the area under the curve (AUC), optimal UGSR threshold, sensitivity and specificity in differentiating PTMCs and BMNs in the two groups. Results The UGSR of PTMC and BMN was 0.52 ± 0.12 and 0.85 ± 0.24 in the HT group ( P  < 0.001), and 0.57 ± 0.13 and 0.87 ± 0.20 in the non-HT group ( P  < 0.001), respectively. The difference in PTMC-UGSR was significant between the two groups ( P  < 0.001), whereas BMN-UGSR did not differ between the two groups ( P  = 0.416). The AUC, optimal UGSR threshold, sensitivity and specificity of UGSR for differentiating PTMC and BMN in the HT and non-HT group were 0.890 versus 0.901, 0.68 versus 0.72, 91.23% versus 90.67%, and 77.46% versus 82.05%, respectively. Conclusions The USGR of the HT group was lower than that of the non-HT group. Moreover, UGSR exhibited important diagnostic value in differentiating PTMC from BMN in both HT and non-HT groups.

Retrieval-augmented generation (RAG) is usually integrated into large language models (LLMs) to mitigate hallucinations and knowledge obsolescence. Whereas,conventional one-step retrieve-and-read methods are insufficient for multi-hop question answering, facing challenges of retrieval semantic mismatching and the high cost in handling interdependent subquestions. In this paper, we propose Optimizing Question Semantic Space for Dynamic Retrieval-Augmented Multi-hop Question Answering (Q-DREAM). Q-DREAM consists of three key modules: (1) the Question Decomposition Module (QDM), which decomposes multi-hop questions into fine-grained subquestions; (2) the Subquestion Dependency Optimizer Module (SDOM), which models the interdependent relations of subquestions for better understanding; and (3) the Dynamic Passage Retrieval Module (DPRM), which aligns subquestions with relevant passages by optimizing the semantic embeddings. Experimental results across various benchmarks demonstrate that Q-DREAM significantly outperforms existing RAG methods, achieving state-of-the-art performance in both in-domain and out-of-domain settings. Notably, Q-DREAM also improves retrieval efficiency while maintaining high accuracy compared with recent baselines.

As Large Language Models increasingly automate complex, long-horizon tasks such as \\emph{vibe coding}, a supervision gap has emerged. While models excel at execution, users often struggle to guide them effectively due to insufficient domain expertise, the difficulty of articulating precise intent, and the inability to reliably validate complex outputs. It presents a critical challenge in scalable oversight: enabling humans to responsibly steer AI systems on tasks that surpass their own ability to specify or verify. To tackle this, we propose Scalable Interactive Oversight, a framework that decomposes complex intent into a recursive tree of manageable decisions to amplify human supervision. Rather than relying on open-ended prompting, our system elicits low-burden feedback at each node and recursively aggregates these signals into precise global guidance. Validated in web development task, our framework enables non-experts to produce expert-level Product Requirement Documents, achieving a 54\\% improvement in alignment. Crucially, we demonstrate that this framework can be optimized via Reinforcement Learning using only online user feedback, offering a practical pathway for maintaining human control as AI scales.

Large language models make remarkable progress in reasoning capabilities. Existing works focus mainly on deductive reasoning tasks (e.g., code and math), while another type of reasoning mode that better aligns with human learning, inductive reasoning, is not well studied. We attribute the reason to the fact that obtaining high-quality process supervision data is challenging for inductive reasoning. Towards this end, we novelly employ number sequences as the source of inductive reasoning data. We package sequences into algorithmic problems to find the general term of each sequence through a code solution. In this way, we can verify whether the code solution holds for any term in the current sequence, and inject case-based supervision signals by using code unit tests. We build a sequence synthetic data pipeline and form a training dataset CodeSeq. Experimental results show that the models tuned with CodeSeq improve on both code and comprehensive reasoning benchmarks.

The evolution of Large Language Models (LLMs) into autonomous agents has expanded the scope of AI coding from localized code generation to complex, repository-level, and execution-driven problem solving. However, current benchmarks predominantly evaluate code logic in static contexts, neglecting the dynamic, full-process requirements of real-world engineering, particularly in backend development which demands rigorous environment configuration and service deployment. To address this gap, we introduce ABC-Bench, a benchmark explicitly designed to evaluate agentic backend coding within a realistic, executable workflow. Using a scalable automated pipeline, we curated 224 practical tasks spanning 8 languages and 19 frameworks from open-source repositories. Distinct from previous evaluations, ABC-Bench require the agents to manage the entire development lifecycle from repository exploration to instantiating containerized services and pass the external end-to-end API tests. Our extensive evaluation reveals that even state-of-the-art models struggle to deliver reliable performance on these holistic tasks, highlighting a substantial disparity between current model capabilities and the demands of practical backend engineering. Our code is available at https://github.com/OpenMOSS/ABC-Bench.

Retrieval-Augmented Generation (RAG) aims to generate more reliable and accurate responses, by augmenting large language models (LLMs) with the external vast and dynamic knowledge. Most previous work focuses on using RAG for single-round question answering, while how to adapt RAG to the complex conversational setting wherein the question is interdependent on the preceding context is not well studied. In this paper, we propose a conversation-level RAG approach, which incorporates fine-grained retrieval augmentation and self-check for conversational question answering (CQA). In particular, our approach consists of three components, namely conversational question refiner, fine-grained retriever and self-check based response generator, which work collaboratively for question understanding and relevant information acquisition in conversational settings. Extensive experiments demonstrate the great advantages of our approach over the state-of-the-art baselines. Moreover, we also release a Chinese CQA dataset with new features including reformulated question, extracted keyword, retrieved paragraphs and their helpfulness, which facilitates further researches in RAG enhanced CQA.