A5101571008- Intermetallics and Advanced Alloy Properties
- Advanced materials and composites
- Molten salt chemistry and electrochemical processes
- Advanced ceramic materials synthesis
- Hydrology and Watershed Management Studies
- Aluminum Alloys Composites Properties
- Natural Language Processing Techniques
- Metallurgical Processes and Thermodynamics
- Metallurgical and Alloy Processes
- Advanced battery technologies research
- Topic Modeling
- Industrial Vision Systems and Defect Detection
- RFID technology advancements
- Soil erosion and sediment transport
- Bauxite Residue and Utilization
- Pigment Synthesis and Properties
- Environmental Changes in China
- Domain Adaptation and Few-Shot Learning
- Digital and Cyber Forensics
- Advanced Image Fusion Techniques
- Machine Learning and ELM
- Advancements in Solid Oxide Fuel Cells
- Microwave-Assisted Synthesis and Applications
- Climate variability and models
- Nuclear materials and radiation effects
Anhui University of Science and Technology
2020-2024
University of Science and Technology Beijing
2011-2022
Jilin University
2020
Harbin Institute of Technology
2019
Beijing Normal University
2017
China Academy of Railway Sciences
2016
China Three Gorges University
2014
Heilongjiang Academy of Land Reclamation Sciences
2012
National Taiwan Ocean University
2005
Recent advances in large language models (LLMs) provide a new opportunity to address the long-standing question of how concepts are represented and organized mind, which is central unravelling nature human cognition. Here, we reframed classic reverse dictionary task simulate concept inference context investigated emergence human-like conceptual representations within LLMs. We found that LLMs were able infer from definitional descriptions construct representation spaces converge towards...
Surface treatment is known as a very efficient measure by which to modulate the surface properties of biomaterials in terms grain structure, topography, roughness and chemistry determine osseointegration implants. In this work, two-step method modification was employed impart high osteogenic activity biomineralization capacity on Ti-25Nb-3Mo-2Sn-3Zr alloy (a type β-titanium named TLM). The preliminary mechanical attrition (SMAT) refined average size from 170 ± 19 μm 74 8 nm TLM layer...
Large language models (LLMs) are being rapidly developed, and a key component of their widespread deployment is safety-related alignment. Many red-teaming efforts aim to jailbreak LLMs, where among these efforts, the Greedy Coordinate Gradient (GCG) attack's success has led growing interest in study optimization-based jailbreaking techniques. Although GCG significant milestone, its attacking efficiency remains unsatisfactory. In this paper, we present several improved (empirical) techniques...
Fine Nb–Sn and Nb–Ta intermetallic powders were produced by hydrogen reduction from their uniformly mixed chloride vapours, while Nb–Al synthesised sodium of chlorides in a LiCl–KCl–NaCl–CaCl2 molten medium. The obtained fine agglomerated, with uniform particle size distribution typical 20–40 nm. was approximately the range 50–300
Ground penetrating radar (GPR) velocity tomography has shown much promise in the detection of geological hazardous bodies. Using GPR technology, faults parallel to transmission direction can be more accurately detected than perpendicular direction. This indicates that ray coverage a great influence on inversion accuracy. In order identify location, scope, and characteristics disaster areas, an image fusion strategy based wavelet transform was proposed because it better multi-resolution...
As one of the most effective self-supervised representation learning methods, contrastive (CL) relies on multiple negative pairs to contrast against each positive pair. In standard practice learning, data augmentation methods are utilized generate both and pairs. While existing works have been focusing improving sampling, sampling process is often overlooked. fact, generated samples polluted by samples, which leads a biased loss performance degradation. To correct bias, we propose novel...
Large language models (LLMs) may memorize sensitive or copyrighted content, raising privacy and legal concerns. Due to the high cost of retraining from scratch, researchers attempt employ machine unlearning remove specific content LLMs while preserving overall performance. In this paper, we discuss several issues in for provide our insights on possible approaches. To address issue inadequate evaluation model outputs after unlearning, introduce three additional metrics evaluate token...
The research on geological landscape has received more and attention worldwide. National Geological Park of Qian-an mud forest, located in Country, Songyuan City (Jilin Province, China) is a rare natural formed by erosion. Mud forest undergone long-term processes, it still continuous evolution due to subsurface In the process formation evolution, distinct stages have been recognized. erosion factors area were identified groundwater soil samples characterization, mechanism studied. Results...
We reported a structure-based virtual screening of acompound library derived from 3-acyl-5-hydroxybenzofurans to develop potentialdrugs for treating breast cancer. A 160,000 compounds was generatedand screened based on the G-score between ligands and receptor ERα. The topstructures were further analyzed evaluate receptor-ligand bindinginteractions. By comparing binding characteristics docking scoringvalues existing ERα analogues, we determined top 200 aspotential drug candidates
Through the double mass curves analysis, it concluded that relationship between areal and runoff depth within upper reaches of Naoli River, River basin had experienced a great change since 1963, especially coefficient which appeared in decreasing trend. By adopting difference Rank T test (α=0.05,n1=7,n2=43,T=255,T1α=108.52,T2α=248.47, T＞T2α＞T1α), also there was no consistency series from 1956 to 1962 ones 1963 2005. analyzing frequency precipitation-runoff depth, indicated that, under same...