Yixuan Liu

Tong Xie, Yuwei Wan, Wei Huang et al.

Emerging tools bring forth fresh approaches to work, and the field of natural science is no different. In natural science, traditional manual, serial, and labour-intensive work is being augmented by automated, parallel, and iterative processes driven by artificial intelligence-based experimental automation and more. To add new capabilities in natural science, enabling the acceleration and enrichment of automation of the discovery process, we present DARWIN, a series of tailored LLMs for natural science, mainly in physics, chemistry, and material science. This series relies on open-source LLM, incorporating structured and unstructured scientific knowledge from public datasets and literature. We fine-tuned the models using over 60,000 instruction data points, emphasizing factual correctness. During the fine-tuning, we introduce the Scientific Instruction Generation (SIG) model, automating instruction generation from scientific texts. This eliminates the need for manual extraction or domain-specific knowledge graphs and efficiently injects scientific knowledge into the model. We also explore multi-task training strategies, revealing interconnections between scientific tasks. DARWIN series not only achieves state-of-the-art results on various scientific tasks but also diminishes reliance on closed-source AI models. Our research showcases the ability of LLM in the scientific domain, with the overarching goal of fostering prosperity within the broader AI for science community.

Jinyu Guo, Kai Shuang, Jijie Li et al.

In dialogue state tracking, dialogue history is a crucial material, and its utilization varies between different models. However, no matter how the dialogue history is used, each existing model uses its own consistent dialogue history during the entire state tracking process, regardless of which slot is updated. Apparently, it requires different dialogue history to update different slots in different turns. Therefore, using consistent dialogue contents may lead to insufficient or redundant information for different slots, which affects the overall performance. To address this problem, we devise DiCoS-DST to dynamically select the relevant dialogue contents corresponding to each slot for state updating. Specifically, it first retrieves turn-level utterances of dialogue history and evaluates their relevance to the slot from a combination of three perspectives: (1) its explicit connection to the slot name; (2) its relevance to the current turn dialogue; (3) Implicit Mention Oriented Reasoning. Then these perspectives are combined to yield a decision, and only the selected dialogue contents are fed into State Generator, which explicitly minimizes the distracting information passed to the downstream state prediction. Experimental results show that our approach achieves new state-of-the-art performance on MultiWOZ 2.1 and MultiWOZ 2.2, and achieves superior performance on multiple mainstream benchmark datasets (including Sim-M, Sim-R, and DSTC2).

Jie Liu, Yixuan Liu, Xue Han et al.

Previous contrastive learning methods for sentence representations often focus on insensitive transformations to produce positive pairs, but neglect the role of sensitive transformations that are harmful to semantic representations. Therefore, we propose an Equivariant Self-Contrastive Learning (ESCL) method to make full use of sensitive transformations, which encourages the learned representations to be sensitive to certain types of transformations with an additional equivariant learning task. Meanwhile, in order to improve practicability and generality, ESCL simplifies the implementations of traditional equivariant contrastive methods to share model parameters from the perspective of multi-task learning. We evaluate our ESCL on semantic textual similarity tasks. The proposed method achieves better results while using fewer learning parameters compared to previous methods.

Tong Xie, Yuwei Wan, Yixuan Liu et al.

Materials discovery and design aim to find compositions and structures with desirable properties over highly complex and diverse physical spaces. Traditional solutions, such as high-throughput simulations or machine learning, often rely on complex descriptors, which hinder generalizability and transferability across different material systems. Moreover, These descriptors may inadequately represent macro-scale material properties, which are influenced by structural imperfections and compositional variations in real-world samples, thus limiting their practical applicability. To address these challenges, we propose DARWIN 1.5, the largest open-source large language model tailored for materials science. By leveraging natural language as input, DARWIN eliminates the need for task-specific descriptors and enables a flexible, unified approach to material property prediction and discovery. Our approach integrates 6M material domain papers and 21 experimental datasets from 49,256 materials across modalities while enabling cross-task knowledge transfer. The enhanced model achieves up to 59.1% improvement in prediction accuracy over the base LLaMA-7B architecture and outperforms SOTA machine learning approaches across 8 materials design tasks. These results establish LLMs as a promising foundation for developing versatile and scalable models in materials science.

Yuwei Wan, Yixuan Liu, Aswathy Ajith et al.

We introduce SciQAG, a novel framework for automatically generating high-quality science question-answer pairs from a large corpus of scientific literature based on large language models (LLMs). SciQAG consists of a QA generator and a QA evaluator, which work together to extract diverse and research-level questions and answers from scientific papers. Utilizing this framework, we construct a large-scale, high-quality, open-ended science QA dataset containing 188,042 QA pairs extracted from 22,743 scientific papers across 24 scientific domains. We also introduce SciQAG-24D, a new benchmark task designed to evaluate the science question-answering ability of LLMs. Extensive experiments demonstrate that fine-tuning LLMs on the SciQAG dataset significantly improves their performance on both open-ended question answering and scientific tasks. To foster research and collaboration, we make the datasets, models, and evaluation codes publicly available, contributing to the advancement of science question answering and developing more interpretable and reasoning-capable AI systems.

Haichao Sha, Ruixuan Liu, Yixuan Liu et al.

The paradigm of Differentially Private SGD~(DP-SGD) can provide a theoretical guarantee for training data in both centralized and federated settings. However, the utility degradation caused by DP-SGD limits its wide application in high-stakes tasks, such as medical image diagnosis. In addition to the necessary perturbation, the convergence issue is attributed to the information loss on the gradient clipping. In this work, we propose a general framework PCDP-SGD, which aims to compress redundant gradient norms and preserve more crucial top gradient components via projection operation before gradient clipping. Additionally, we extend PCDP-SGD as a fundamental component in differential privacy federated learning~(DPFL) for mitigating the data heterogeneous challenge and achieving efficient communication. We prove that pre-projection enhances the convergence of DP-SGD by reducing the dependence of clipping error and bias to a fraction of the top gradient eigenspace, and in theory, limits cross-client variance to improve the convergence under heterogeneous federation. Experimental results demonstrate that PCDP-SGD achieves higher accuracy compared with state-of-the-art DP-SGD variants in computer vision tasks. Moreover, PCDP-SGD outperforms current federated learning frameworks when DP is guaranteed on local training sets.

Yixuan Liu, Chrysafis Vogiatzis, Ruriko Yoshida et al.

Uncrewed autonomous vehicles (UAVs) have made significant contributions to reconnaissance and surveillance missions in past US military campaigns. As the prevalence of UAVs increases, there has also been improvements in counter-UAV technology that makes it difficult for them to successfully obtain valuable intelligence within an area of interest. Hence, it has become important that modern UAVs can accomplish their missions while maximizing their chances of survival. In this work, we specifically study the problem of identifying a short path from a designated start to a goal, while collecting all rewards and avoiding adversaries that move randomly on the grid. We also provide a possible application of the framework in a military setting, that of autonomous casualty evacuation. We present a comparison of three methods to solve this problem: namely we implement a Deep Q-Learning model, an $\varepsilon$-greedy tabular Q-Learning model, and an online optimization framework. Our computational experiments, designed using simple grid-world environments with random adversaries showcase how these approaches work and compare them in terms of performance, accuracy, and computational time.

Yixuan Liu, Hu Wang, Xiaowei Wang et al.

Appropriate credit assignment for delay rewards is a fundamental challenge for reinforcement learning. To tackle this problem, we introduce a delay reward calibration paradigm inspired from a classification perspective. We hypothesize that well-represented state vectors share similarities with each other since they contain the same or equivalent essential information. To this end, we define an empirical sufficient distribution, where the state vectors within the distribution will lead agents to environmental reward signals in the consequent steps. Therefore, a purify-trained classifier is designed to obtain the distribution and generate the calibrated rewards. We examine the correctness of sufficient state extraction by tracking the real-time extraction and building different reward functions in environments. The results demonstrate that the classifier could generate timely and accurate calibrated rewards. Moreover, the rewards are able to make the model training process more efficient. Finally, we identify and discuss that the sufficient states extracted by our model resonate with the observations of humans.