A5067986384- Cancer Immunotherapy and Biomarkers
- Cutaneous Melanoma Detection and Management
- Immunotherapy and Immune Responses
- Topic Modeling
- Natural Language Processing Techniques
- Advancements in Battery Materials
- Advanced Battery Materials and Technologies
- Advanced Battery Technologies Research
- Multimodal Machine Learning Applications
- Lung Cancer Treatments and Mutations
- Radiomics and Machine Learning in Medical Imaging
- Epigenetics and DNA Methylation
- Expert finding and Q&A systems
- HER2/EGFR in Cancer Research
- Cholangiocarcinoma and Gallbladder Cancer Studies
- Advanced Text Analysis Techniques
- Extraction and Separation Processes
- Monoclonal and Polyclonal Antibodies Research
- Environmental remediation with nanomaterials
- Sentiment Analysis and Opinion Mining
- Breast Cancer Treatment Studies
- Esophageal Cancer Research and Treatment
- Advanced Graph Neural Networks
- Advanced oxidation water treatment
- Inorganic Chemistry and Materials
Argonne National Laboratory
2023-2025
Chinese Academy of Medical Sciences & Peking Union Medical College
2015-2024
University of Connecticut
2023-2024
Sohu (China)
2019-2021
Max Planck Society
2020
Fudan University
2020
Qingdao University
2019
Chinese Academy of Sciences
2016-2019
National Clinical Research
2019
University of Chinese Academy of Sciences
2017-2018
sentences (Hermann et al., 2015;Rocktäschel 2015;Tan 2015). Based on recurrent neural networks (RNN), external attention information was added to hidden representations get an attentive sentence representation.Despite the improvement over nonattentive models, mechanism under RNN is not well studied.In this work, we analyze deficiency of traditional based models quantitatively and qualitatively.Then present three new that add before representation, which shows advantage in representing...
As Large Language Models (LLMs) continue to advance in performance, their size has escalated significantly, with current LLMs containing billions or even trillions of parameters. However, this study, we discovered that many layers exhibit high similarity, and some play a negligible role network functionality. Based on observation, define metric called Block Influence (BI) gauge the significance each layer LLMs. We then propose straightforward pruning approach: removal, which directly delete...
Multivalent-ion battery technologies are increasingly attractive options for meeting diverse energy storage needs. Calcium ion batteries (CIB) particularly appealing candidates their earthly abundance, high theoretical volumetric density, and relative safety advantages. At present, only a few Ca-ion electrolyte systems reported to reversibly plate at room temperature: example, aluminates borates, including Ca[TPFA]2, where [TPFA]− = [Al(OC(CF3)3)4]− Ca[B(hfip)4]2, [B(hfip)4]2–...
Recently proposed Story Cloze Test [Mostafazadeh et al., 2016] is a commonsense machine comprehension application to deal with natural language understanding problem. This dataset contains lot of story tests which require inference ability. Unfortunately, the training data almost unsupervised where each context document followed only one positive sentence that can be inferred from context. However, in testing period, we must make two candidate sentences. To tackle this problem, employ...
Abstract Objective: The novel fully automated immunohistochemistry (IHC) assay-Ventana anaplastic lymphoma kinase (ALK)-D5F3 for screening ALK rearrangements has been approved by China's Food and Drug Administration in 2013, our previous study disclosed a highly specificity sensitivity nearly 100%, its efficacy needs to be evaluated large cohort of primary lung adenocarcinoma patients, compare clinicopathological features with (+) (-) adenocarcinoma. Methods: A total 1,504 consecutive...
Large language models (LLMs) have gained extended context windows through scaling positional encodings and lightweight continual pre-training. However, this often leads to degraded performance on short-text tasks, while the reasons for degradation remain insufficiently explored. In work, we identify two primary factors contributing issue: distribution drift in hidden states attention scores, catastrophic forgetting during To address these challenges, propose Long Context Pre-training with...
The current generation of large language models (LLMs) is typically designed for broad, general-purpose applications, while domain-specific LLMs, especially in vertical fields like medicine, remain relatively scarce. In particular, the development highly efficient and practical LLMs medical domain challenging due to complexity knowledge limited availability high-quality data. To bridge this gap, we introduce Baichuan-M1, a series specifically optimized applications. Unlike traditional...
Aspect-based sentiment analysis (ABSA) aims to predict the towards a specific aspect in text. However, existing ABSA test sets cannot be used probe whether model can distinguish of target from non-target aspects. To solve this problem, we develop simple but effective approach enrich sets. Specifically, generate new examples disentangle confounding sentiments aspects aspect’s sentiment. Based on SemEval 2014 dataset, construct Aspect Robustness Test Set (ARTS) as comprehensive robustness...
Abstract Aim Pathologists are currently supposed to be aware of both domestic and international guidelines for breast cancer diagnosis, but it is unclear how successfully these have been integrated into routine clinical practice in China. Thus, this national proficiency testing (PT) scheme pathology was set up conduct a baseline assessment the diagnostic capability pathologists Methods This PT plan designed implemented according “Conformity assessment—General requirements testing”...
Neural question generation (NQG) is the task of generating questions from given context with deep neural networks. Previous answer-aware NQG methods suffer problem that generated answers are focusing on entity and most trivial to be answered. The answer-agnostic reduce bias towards named entities increasing model's degrees freedom, but sometimes result in unanswerable which not valuable for subsequent machine reading comprehension system. In this paper, we treat as hidden pivot combine...
To determine the prevalence and clinicopathological characteristics of BRAF V600E mutation HER2 exon 20 insertions in Chinese lung adenocarcinoma (ADC) patients.Given fact that driver mutations are mutually exclusive ADCs, 204 EGFR/KRAS wild-type cases were enrolled this study. Direct Sanger sequencing was performed to examine mutations. The association with statistically analyzed.Among ADCs tested, 11 (5.4%) carried 4 (2.0%) had mutation. status identified be associated a non-smoking...
Open-domain question answering focuses on using diverse information resources to answer any types of question. Recent years, with the development large-scale data set and various deep neural networks models, some recent advances in open domain system first utilize distantly supervised dataset as knowledge resource, then apply learning based machine comprehension techniques generate right answers, which achieves impressive results compared traditional feature-based pipeline methods.
This paper presents the ReCO, a human-curated Chinese Reading Comprehension dataset on Opinion. The questions in ReCO are opinion based queries issued to commercial search engine. passages provided by crowdworkers who extract support snippet from retrieved documents. Finally, an abstractive yes/no/uncertain answer was given crowdworkers. release of consists 300k that our knowledge is largest reading comprehension. A prominent characteristic addition original context paragraph, we also...
Despite great progress in surgery and other treatments, the prognosis of patients with esophageal squamous cell carcinoma (ESCC) is still very poor. HER2 has strong therapeutic implications certain cancers, such as breast cancer gastric cancer. However, literature on frequency expression ESCC scarce. In present study, protein expression, gene amplification relationship between status clinicopathological characteristics were evaluated a large cohort Chinese patients.A total 857 consecutive...
With an increasing demand for intermittent renewable energy and electric vehicles, it is imperative to develop lithium-ion batteries with Earth-abundant cathode materials. Cobalt (Co) preferred be kept at a minimum because of its high cost limited mining options, yet has played essential role in the high-performance transition metal oxides (TMOs). Herein, we report work from Argonne National Laboratory, conducted under U.S. DoE’s Vehicle Technologies Office, Deep Dive consortium on...
Abstract Mucosal melanoma exhibits limited responsiveness to anti-PD-1 therapy. However, a subgroup of mucosal melanomas, particularly those situated at specific anatomic sites like primary malignant the esophagus (PMME), display remarkable sensitivity treatment. The underlying mechanisms driving this superior response and DNA methylation patterns in have not been thoroughly investigated. We collected tumor samples from 50 patients with melanoma, including 31 PMME 19 non-esophageal (NEMM)....