Search

Gradient-free prompt optimization methods have made significant strides in enhancing the performance of closed-source Large Language Models (LLMs) across a wide range of tasks. However, existing approaches make light of the importance of high-quality prompt initialization and the identification of effective optimization directions, thus resulting in substantial optimization steps to obtain satisfactory performance. In this light, we aim to accelerate prompt optimization process to tackle the challenge of low convergence rate. We propose a dual-phase approach which starts with generating high-quality initial prompts by adopting a well-designed meta-instruction to delve into task-specific information, and iteratively optimize the prompts at the sentence level, leveraging previous tuning experience to expand prompt candidates and accept effective ones. Extensive experiments on eight datasets demonstrate the effectiveness of our proposed method, achieving a consistent accuracy gain over baselines with less than five optimization steps., Comment: EMNLP 2024 Findings

In this work, rate coefficients of four prototypical insertion reactions, X + H2 -- H + XH (X=C(1D), N(2D), O(1D), S(1D)) and associated isotope reactions are calculated based on ring polymer molecular dynamics (RPMD) with Cayley propagator (Cayley-RPMD). The associated kinetic isotope effects (KIEs) are systematically studied too. The Cayley propagator used in this work increases the stability of numerical integration in RPMD calculations, and also supports a larger evolution time interval, allowing us to reach both high accuracy and efficiency. So, our results do not only provide chemical kinetic data for the title reactions in an extended temperature range, but also consist of experimental results, standard RPMD, and other theoretical methods. The results in this work also reflect that Cayley-RPMD has strong consistency and high reliability in the investigations of chemical dynamics for insertion reactions., Comment: 19 pages, 4 figures

Globally accurate full-dimensional ground state potential energy surface (PES) for the Cl($^2$P) + XCl $\to$ HCl + Cl($^2$P) reaction, a prototypical heavy-light-heavy abstract reaction, is developed using permutation invariant polynomial neural network (PIP-NN) method and embedded atom neural network (EANN) method, with the corresponding total root mean square error (RMSE) being only 0.043 and 0.056 kcal/mol, respectively. The saddle point of this reaction system is found to be nonlinear. A full-dimensional approximate quantum mechanical method, ring-polymer molecular dynamics (RPMD) with Cayley propagator, is employed to calculate the thermal rate coefficients and kinetic isotopic effects of title reactions Cl($^2$P) + XCl $\to$ XCl + Cl($^2$P) (X = H, D, Mu) on both new PESs. The results reproduce the experimental results at high temperatures perfectly, but with moderate accuracy at lower temperatures. The similar kinetic behavior is supported by quantum dynamics using wave packet calculations as well., Comment: 23 pages,5 figures

The force field describing the calculated interaction between atoms or molecules is the key to the accuracy of many molecular dynamics (MD) simulation results. Compared with traditional or semi-empirical force fields, machine learning force fields have the advantages of faster speed and higher precision. We have employed the method of atomic cluster expansion (ACE) combined with first-principles density functional theory (DFT) calculations for machine learning, and successfully obtained the force field of the binary Fe-Co alloy. Molecular dynamics simulations of Fe-Co alloy carried out using this ACE force field predicted the correct phase transition range of Fe-Co alloy., Comment: 17 pages, 6 figures

The accurate and efficient calculation of the rate coefficients of chemical reactions is a key issue in the research of chemical dynamics. In this work, by applying the dimension-free ultra-stable Cayley propagator, the thermal rate coefficients of a prototypic high dimensional chemical reaction OH + CH4 -> H2O + CH3 in the temperature range of 200 K to 1500 K are investigated with ring polymer molecular dynamics (RPMD), on a highly accurate full-dimensional potential energy surface. Kinetic isotope effects (KIEs) for three isotopologues of the title reaction are also studied. The results demonstrate excellent agreement with experimental data, even in the deep tunneling region. Especially, the Cayley propagator shows high calculation efficiency with little loss of accuracy. The present results confirmed the applicability of the RPMD method, particularly the speed-up by Cayley propagator, in theoretical calculations of bimolecular reaction rates., Comment: 35 pages, 7 figures (include one TOC)

Molecular ferroelectrics have captured immense attention due to their superiority over inorganic oxide ferroelectrics, such as environmentally friendly, low-cost, flexible, foldable. However, the mechanisms of ferroelectric switching and phase transition for the molecular ferroelectrics are still missing, leaving the development of novel molecular ferroelectrics less efficient. In this work, we have provided a methodology combining molecular dynamics (MD) simulation on a polarized force field named polarized crystal charge (PCC) and enhanced sampling technique, replica-exchange molecular dynamics (REMD) to simulate such mechanisms. With this procedure, we have investigated a promising molecular ferroelectric material, (R)/(S)-3-quinuclidinol crystal. We have simulated the ferroelectric hysteresis loops of both enantiomers and obtained spontaneous polarization of 7/8 \mu C cm-2 and a corresponding coercive electric field of 15 kV cm-1. We also find the Curie temperature as 380/385 K for ferro-/para-electric phase transition of both enantiomers. All of the simulated results are highly compatible with experimental values. Besides of that, we predict a novel Curie temperature of about 600 K. This finding is further validated by principal component analysis (PCA). Our work would significantly promote the future exploration of multifunctional molecular ferroelectrics for the next generation of intelligent devices.