Search

This work introduces Weaver, our first family of large language models (LLMs) dedicated to content creation. Weaver is pre-trained on a carefully selected corpus that focuses on improving the writing capabilities of large language models. We then fine-tune Weaver for creative and professional writing purposes and align it to the preference of professional writers using a suit of novel methods for instruction data synthesis and LLM alignment, making it able to produce more human-like texts and follow more diverse instructions for content creation. The Weaver family consists of models of Weaver Mini (1.8B), Weaver Base (6B), Weaver Pro (14B), and Weaver Ultra (34B) sizes, suitable for different applications and can be dynamically dispatched by a routing agent according to query complexity to balance response quality and computation cost. Evaluation on a carefully curated benchmark for assessing the writing capabilities of LLMs shows Weaver models of all sizes outperform generalist LLMs several times larger than them. Notably, our most-capable Weaver Ultra model surpasses GPT-4, a state-of-the-art generalist LLM, on various writing scenarios, demonstrating the advantage of training specialized LLMs for writing purposes. Moreover, Weaver natively supports retrieval-augmented generation (RAG) and function calling (tool usage). We present various use cases of these abilities for improving AI-assisted writing systems, including integration of external knowledge bases, tools, or APIs, and providing personalized writing assistance. Furthermore, we discuss and summarize a guideline and best practices for pre-training and fine-tuning domain-specific LLMs.

Medical image segmentation aims to identify and locate abnormal structures in medical images, such as chest radiographs, using deep neural networks. These networks require a large number of annotated images with fine-grained masks for the regions of interest, making pre-training strategies based on classification datasets essential for sample efficiency. Based on a large-scale medical image classification dataset, our work collects explanations from well-trained classifiers to generate pseudo labels of segmentation tasks. Specifically, we offer a case study on chest radiographs and train image classifiers on the CheXpert dataset to identify 14 pathological observations in radiology. We then use Integrated Gradients (IG) method to distill and boost the explanations obtained from the classifiers, generating massive diagnosis-oriented localization labels (DoLL). These DoLL-annotated images are used for pre-training the model before fine-tuning it for downstream segmentation tasks, including COVID-19 infectious areas, lungs, heart, and clavicles. Our method outperforms other baselines, showcasing significant advantages in model performance and training efficiency across various segmentation settings.

Given the complexity and lack of transparency in deep neural networks (DNNs), extensive efforts have been made to make these systems more interpretable or explain their behaviors in accessible terms. Unlike most reviews, which focus on algorithmic and model-centric perspectives, this work takes a "data-centric" view, examining how data collection, processing, and analysis contribute to explainable AI (XAI). We categorize existing work into three categories subject to their purposes: interpretations of deep models, referring to feature attributions and reasoning processes that correlate data points with model outputs; influences of training data, examining the impact of training data nuances, such as data valuation and sample anomalies, on decision-making processes; and insights of domain knowledge, discovering latent patterns and fostering new knowledge from data and models to advance social values and scientific discovery. Specifically, we distill XAI methodologies into data mining operations on training and testing data across modalities, such as images, text, and tabular data, as well as on training logs, checkpoints, models and other DNN behavior descriptors. In this way, our study offers a comprehensive, data-centric examination of XAI from a lens of data mining methods and applications.

Editing objects within a scene is a critical functionality required across a broad spectrum of applications in computer vision and graphics. As 3D Gaussian Splatting (3DGS) emerges as a frontier in scene representation, the effective modification of 3D Gaussian scenes has become increasingly vital. This process entails accurately retrieve the target objects and subsequently performing modifications based on instructions. Though available in pieces, existing techniques mainly embed sparse semantics into Gaussians for retrieval, and rely on an iterative dataset update paradigm for editing, leading to over-smoothing or inconsistency issues. To this end, this paper proposes a systematic approach, namely TIGER, for coherent text-instructed 3D Gaussian retrieval and editing. In contrast to the top-down language grounding approach for 3D Gaussians, we adopt a bottom-up language aggregation strategy to generate a denser language embedded 3D Gaussians that supports open-vocabulary retrieval. To overcome the over-smoothing and inconsistency issues in editing, we propose a Coherent Score Distillation (CSD) that aggregates a 2D image editing diffusion model and a multi-view diffusion model for score distillation, producing multi-view consistent editing with much finer details. In various experiments, we demonstrate that our TIGER is able to accomplish more consistent and realistic edits than prior work.

Context.Identification of new star cluster candidates in M31 is fundamental for the study of the M31 stellar cluster system. The machine-learning method convolutional neural network (CNN) is an efficient algorithm for searching for new M31 star cluster candidates from tens of millions of images from wide-field photometric surveys. Aims.We search for new M31 cluster candidates from the high-quality $g$- and $i$-band images of 21,245,632 sources obtained from the Pan-Andromeda Archaeological Survey (PAndAS) through a CNN. Methods.We collected confirmed M31 clusters and noncluster objects from the literature as our training sample. Accurate double-channel CNNs were constructed and trained using the training samples. We applied the CNN classification models to the PAndAS $g$- and $i$-band images of over 21 million sources to search new M31 cluster candidates. The CNN predictions were finally checked by five experienced human inspectors to obtain high-confidence M31 star cluster candidates. Results.After the inspection, we identified a catalogue of 117 new M31 cluster candidates. Most of the new candidates are young clusters that are located in the M31 disk. Their morphology, colours, and magnitudes are similar to those of the confirmed young disk clusters. We also identified eight globular cluster candidates that are located in the M31 halo and exhibit features similar to those of confirmed halo globular clusters. The projected distances to the M31 centre for three of them are larger than 100\,kpc., Comment: Published in A&A

Graph neural architecture search has received a lot of attention as Graph Neural Networks (GNNs) has been successfully applied on the non-Euclidean data recently. However, exploring all possible GNNs architectures in the huge search space is too time-consuming or impossible for big graph data. In this paper, we propose a parallel graph architecture search (GraphPAS) framework for graph neural networks. In GraphPAS, we explore the search space in parallel by designing a sharing-based evolution learning, which can improve the search efficiency without losing the accuracy. Additionally, architecture information entropy is adopted dynamically for mutation selection probability, which can reduce space exploration. The experimental result shows that GraphPAS outperforms state-of-art models with efficiency and accuracy simultaneously., Comment: 5 papes,3 figures,Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval

In spite of maturity to the modern electronic design automation (EDA) tools, optimized designs at architectural stage may become suboptimal after going through physical design flow. Adder design has been such a long studied fundamental problem in very large-scale integration industry yet designers cannot achieve optimal solutions by running EDA tools on the set of available prefix adder architectures. In this paper, we enhance a state-of-the-art prefix adder synthesis algorithm to obtain a much wider solution space in architectural domain. On top of that, a machine learning-based design space exploration methodology is applied to predict the Pareto frontier of the adders in physical domain, which is infeasible by exhaustively running EDA tools for innumerable architectural solutions. Considering the high cost of obtaining the true values for learning, an active learning algorithm is proposed to select the representative data during learning process, which uses less labeled data while achieving better quality of Pareto frontier. Experimental results demonstrate that our framework can achieve Pareto frontier of high quality over a wide design space, bridging the gap between architectural and physical designs. Source code and data are available at https://github.com/yuzhe630/adder-DSE.

We present a detailed analysis of the beam energy dependence of the mechanisms for the generation of directed and elliptic flows in Au+Au collisions focusing on the role of hadronic rescattering and spectator shadowing within a microscopic transport model JAM with different equation of state. A systematic study of the beam energy dependence is performed for Au+Au collisions at $\sqrt{s_{NN}} =2.3 - 62.4$ GeV. The transition of the dynamical origin of the directed flow is observed. We find that the initial Glauber type nucleon-nucleon collisions generate negative $v_1$ for nucleons at midrapidity due to the presence of spectator matter, and this negative nucleon $v_1$ is turned to be positive by the meson-baryon interactions at the beam energy region of $\sqrt{s_{NN}} < 30$ GeV. In contrast, above 30 GeV there is no spectator shadowing at midrapidity, and initial nucleon-nucleon collisions do not generate directed flow, but subsequent rescatterings among produced particles generate negative $v_1$ for nucleons. It is demonstrated that negative pion-directed flows are mostly generated by the interaction with the spectator matter. It is also shown that the squeeze-out effect is largely suppressed in the case of softening, which leads to the enhancement of elliptic flow around $\sqrt{s_{NN}}=5-7$ GeV. The elliptic flow at midrapidity above 10 GeV is not influenced by the squeeze-out due to spectator matter, while its effect is seen at the forward rapidity range of $y/y_\mathrm{c.m.}>0.5$, which decreases as beam energy increases., Comment: 14 pages, 17 figures

In spite of maturity to the modern electronic design automation (EDA) tools, optimized designs at architectural stage may become sub-optimal after going through physical design flow. Adder design has been such a long studied fundamental problem in VLSI industry yet designers cannot achieve optimal solutions by running EDA tools on the set of available prefix adder architectures. In this paper, we enhance a state-of-the-art prefix adder synthesis algorithm to obtain a much wider solution space in architectural domain. On top of that, a machine learning-based design space exploration methodology is applied to predict the Pareto frontier of the adders in physical domain, which is infeasible by exhaustively running EDA tools for innumerable architectural solutions. Considering the high cost of obtaining the true values for learning, an active learning algorithm is utilized to select the representative data during learning process, which uses less labeled data while achieving better quality of Pareto frontier. Experimental results demonstrate that our framework can achieve Pareto frontier of high quality over a wide design space, bridging the gap between architectural and physical designs., Comment: 14 pages, 15 figures

We perform a systematic study of elliptic flow ($v_2$) in Au+Au collision at $\sqrt{s_{NN}}=5$ GeV by using a microscopic transport model JAM. The centrality, pseudorapidity, transverse momentum and beam energy dependence of $v_2$ for charged as well as identified hadrons are studied. We investigate the effects of both the hadronic mean-field and the softening of equation of state (EoS) on elliptic flow. The softening of EoS is realized by imposing attractive orbits in two body scattering, which can reduce the pressure of the system. We found that the softening of EoS leads to the enhancement of $v_2$, while the hadronic mean-field suppresses $v_2$ relative to the cascade mode. It indicates that elliptic flow at high baryon density regions is highly sensitive to the EoS and the enhancement of $v_2$ may probe the signature of a first-order phase transition in heavy-ion collisions at beam energies of a strong baryon stopping region., Comment: 7 pages, 9 figures, to appear in Chinese Physics C

After more than 20 years of development, China’s capital markets are not yet mature. In order to guide the reform of mature market, the study on the IPO price determination in GEM is necessary. We employ the factor analysis and multiple regression to analyze the issue of IPO pricing on Chinese GEM. Through factor analysis, five factors are extracted which reflect the companies’ intrinsic value. The result of the regression shows that IPO pricing of Chinese GEM is fair. We verify the nonexistence of behavior which lowers the offering price by underwriter. By way of excluding the irrational pricing hypothesis, we conduct the analysis of the IPO underpricing depending on the external factors. The result proves that the market expectation and speculation from investors affect the IPO underpricing most., Includes abstract and appendices., Includes bibliographical references (p. 28-30).