The random distribution problem of dendrite preferred growth direction was settled by random grid method.This method was used to study the influence of forced laminar flow effect on multiple grains during solidificati...The random distribution problem of dendrite preferred growth direction was settled by random grid method.This method was used to study the influence of forced laminar flow effect on multiple grains during solidification.Taking high pure succinonitrile （SCN） undercooled melt as an example,the forced laminar flow effect on multiple grains was studied by phase-field model of single grain which coupled with flow equations at non-isothermal condition.The simulation results show that the random grid method can reasonably settle the problem of random distribution and is more effective.When the solid fraction is relatively low,melt particles flow around the downstream side of dendrite,and the flow velocity between two dendrite arms becomes high.At the stage of solidification time less than 1800Δt,every dendrite grows freely;the upstream dendrites are stronger than the downstream ones.The higher the melt flow rate,the higher the solid fraction.However,when the solid fraction is relatively high,the dendrite arm intertwins and only a little residual melt which is not encapsulated can flow;the solid fraction will gradually tend to equal to solid fraction of melt without flow.展开更多
First-principle simulations have been applied to investigate the effect of copper（Cu） or aluminum（Al） content on the ductility of Al3Ti,AlTi,AlCu,and AlTiCu2 alloys.The mechanical stable and elastic properties of ...First-principle simulations have been applied to investigate the effect of copper（Cu） or aluminum（Al） content on the ductility of Al3Ti,AlTi,AlCu,and AlTiCu2 alloys.The mechanical stable and elastic properties of Al-based intermetallic compounds are researched by density functional theory with the generalized gradient approximation（DFT-GGA）.The calculated lattice constants are in conformity with the previous experimental and theoretical data.The deduced elastic constants show that the investigated Al_3Ti,AlTi,AlCu,and AlTiCu2 structures are mechanically stable.Shear modulus,Young＇s modulus,Poisson＇s ratio,and the ratio B/G have also been figured out by using reckoned elastic constants.A further analysis of Young＇s modulus and Poisson＇s ratio reveals that the third added element copper content has significant effects on the Al-Ti-based ICs ductile character.展开更多
The advancement of Large Language Models (LLMs) for domain applications in fields such as materials science and engineering depends on the development of fine-tuning strategies that adapt models for specialized, techn...The advancement of Large Language Models (LLMs) for domain applications in fields such as materials science and engineering depends on the development of fine-tuning strategies that adapt models for specialized, technical capabilities. In this work, we explore the effects of Continued Pretraining (CPT), Supervised Fine-Tuning (SFT), and various preference-based optimization approaches, including Direct Preference Optimization (DPO) and Odds Ratio Preference Optimization (ORPO), on fine-tuned LLM performance. Our analysis shows how these strategies influence model outcomes and reveals that the merging of multiple fine-tuned models can lead to the emergence of capabilities that surpass the individual contributions of the parent models. We find that model merging is not merely a process of aggregation, but a transformative method that can drive substantial advancements in model capabilities characterized by highly nonlinear interactions between model parameters, resulting in new functionalities that neither parent model could achieve alone, leading to improved performance in domain-specific assessments. We study critical factors that influence the success of model merging, such as the diversity between parent models and the fine-tuning techniques employed. The insights underscore the potential of strategic model merging to unlock novel capabilities in LLMs, offering an effective tool for advancing AI systems to meet complex challenges. Experiments with different model architectures are presented, including the Llama 3.1 8B and Mistral 7B family of models, where similar behaviors are observed. Exploring whether the results hold also for much smaller models, we use a tiny LLM with 1.7 billion parameters and show that very small LLMs do not necessarily feature emergent capabilities under model merging, suggesting that model scaling may be a key component. In open-ended yet consistent chat conversations between a human and AI models, our assessment reveals detailed insights into how different model variants perform, and shows that the smallest model achieves a high intelligence score across key criteria including reasoning depth, creativity, clarity, and quantitative precision. Other experiments include the development of image generation prompts that seek to reason over disparate biological material design concepts, to create new microstructures, architectural concepts, and urban design based on biological materials-inspired construction principles. We conclude with a series of questions about scaling and emergence that could be addressed in future research.展开更多
基金Project(10964004) supported by the National Natural Science Foundation of ChinaProject(20070731001) supported by Research Fund for the Doctoral Program of China+1 种基金 Project(096RJZA104) supported by the Natural Science Foundation of Gansu Province,ChinaProject(SB14200801) supported by the Doctoral Fund of Lanzhou University of Technology,China
文摘The random distribution problem of dendrite preferred growth direction was settled by random grid method.This method was used to study the influence of forced laminar flow effect on multiple grains during solidification.Taking high pure succinonitrile （SCN） undercooled melt as an example,the forced laminar flow effect on multiple grains was studied by phase-field model of single grain which coupled with flow equations at non-isothermal condition.The simulation results show that the random grid method can reasonably settle the problem of random distribution and is more effective.When the solid fraction is relatively low,melt particles flow around the downstream side of dendrite,and the flow velocity between two dendrite arms becomes high.At the stage of solidification time less than 1800Δt,every dendrite grows freely;the upstream dendrites are stronger than the downstream ones.The higher the melt flow rate,the higher the solid fraction.However,when the solid fraction is relatively high,the dendrite arm intertwins and only a little residual melt which is not encapsulated can flow;the solid fraction will gradually tend to equal to solid fraction of melt without flow.
基金Project supported by the National Natural Science Foundation of China(Grant Nos.41674088,11574254,11272296,and 11547311)the National Basic Research Program of China(Grant No.2011CB808201)+1 种基金the Fundamental Research Fund for the Central Universities,China(Grant Nos.2682014ZT30 and 2682014ZT31)the Fund of the State Key Laboratory of Solidification Processing in Northwestern Polytechnical University,China(Grant No.SKLSP201511)
文摘First-principle simulations have been applied to investigate the effect of copper（Cu） or aluminum（Al） content on the ductility of Al3Ti,AlTi,AlCu,and AlTiCu2 alloys.The mechanical stable and elastic properties of Al-based intermetallic compounds are researched by density functional theory with the generalized gradient approximation（DFT-GGA）.The calculated lattice constants are in conformity with the previous experimental and theoretical data.The deduced elastic constants show that the investigated Al_3Ti,AlTi,AlCu,and AlTiCu2 structures are mechanically stable.Shear modulus,Young＇s modulus,Poisson＇s ratio,and the ratio B/G have also been figured out by using reckoned elastic constants.A further analysis of Young＇s modulus and Poisson＇s ratio reveals that the third added element copper content has significant effects on the Al-Ti-based ICs ductile character.
基金supported in part by Google,the MIT Generative AI Initiative,USDA(grant number 2021-69012-35978)with additional support from NIH.This material is partially based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant number 2141064.
文摘The advancement of Large Language Models (LLMs) for domain applications in fields such as materials science and engineering depends on the development of fine-tuning strategies that adapt models for specialized, technical capabilities. In this work, we explore the effects of Continued Pretraining (CPT), Supervised Fine-Tuning (SFT), and various preference-based optimization approaches, including Direct Preference Optimization (DPO) and Odds Ratio Preference Optimization (ORPO), on fine-tuned LLM performance. Our analysis shows how these strategies influence model outcomes and reveals that the merging of multiple fine-tuned models can lead to the emergence of capabilities that surpass the individual contributions of the parent models. We find that model merging is not merely a process of aggregation, but a transformative method that can drive substantial advancements in model capabilities characterized by highly nonlinear interactions between model parameters, resulting in new functionalities that neither parent model could achieve alone, leading to improved performance in domain-specific assessments. We study critical factors that influence the success of model merging, such as the diversity between parent models and the fine-tuning techniques employed. The insights underscore the potential of strategic model merging to unlock novel capabilities in LLMs, offering an effective tool for advancing AI systems to meet complex challenges. Experiments with different model architectures are presented, including the Llama 3.1 8B and Mistral 7B family of models, where similar behaviors are observed. Exploring whether the results hold also for much smaller models, we use a tiny LLM with 1.7 billion parameters and show that very small LLMs do not necessarily feature emergent capabilities under model merging, suggesting that model scaling may be a key component. In open-ended yet consistent chat conversations between a human and AI models, our assessment reveals detailed insights into how different model variants perform, and shows that the smallest model achieves a high intelligence score across key criteria including reasoning depth, creativity, clarity, and quantitative precision. Other experiments include the development of image generation prompts that seek to reason over disparate biological material design concepts, to create new microstructures, architectural concepts, and urban design based on biological materials-inspired construction principles. We conclude with a series of questions about scaling and emergence that could be addressed in future research.
