Parameter efficient finetuning has emerged as a viable solution for improving the performance of Large Language Models without requiring massive resources and compute. Prior work on multilingual evaluation has shown that there is a large gap between the performance of LLMs on English and other languages. Further, there is also a large gap between the performance of smaller open-source models and larger LLMs. Finetuning can be an effective way to bridge this gap and make language models more equitable. In this work, we finetune the LLaMA-7B and Mistral-7B models on synthetic multilingual instruction tuning data to determine its effect on model performance on five downstream tasks covering twenty three languages in all. Additionally, we experiment with various parameters, such as rank for low-rank adaptation and values of quantisation to determine their effects on downstream performance and find that higher rank and higher quantisation values benefit low-resource languages. We find that parameter efficient finetuning of smaller open source models sometimes bridges the gap between the performance of these models and the larger ones, however, English performance can take a hit. We also find that finetuning sometimes improves performance on low-resource languages, while degrading performance on high-resource languages.
相關內容
The growing demand for personalized decision-making has led to a surge of interest in estimating the Conditional Average Treatment Effect (CATE). The intersection of machine learning and causal inference has yielded various effective CATE estimators. However, deploying these estimators in practice is often hindered by the absence of counterfactual labels, making it challenging to select the desirable CATE estimator using conventional model selection procedures like cross-validation. Existing approaches for CATE estimator selection, such as plug-in and pseudo-outcome metrics, face two inherent challenges. Firstly, they are required to determine the metric form and the underlying machine learning models for fitting nuisance parameters or plug-in learners. Secondly, they lack a specific focus on selecting a robust estimator. To address these challenges, this paper introduces a novel approach, the Distributionally Robust Metric (DRM), for CATE estimator selection. The proposed DRM not only eliminates the need to fit additional models but also excels at selecting a robust CATE estimator. Experimental studies demonstrate the efficacy of the DRM method, showcasing its consistent effectiveness in identifying superior estimators while mitigating the risk of selecting inferior ones.
Large Language Models (LLMs) have demonstrated impressive capabilities across a wide range of tasks. However, their proficiency and reliability in the specialized domain of Data Analysis, particularly with a focus on data-driven thinking, remain uncertain. To bridge this gap, we introduce BIBench, a comprehensive benchmark designed to evaluate the data analysis capabilities of LLMs within the context of Business Intelligence (BI). BIBench assesses LLMs across three dimensions: 1) BI foundational knowledge, evaluating the models' numerical reasoning and familiarity with financial concepts; 2) BI knowledge application, determining the models' ability to quickly comprehend textual information and generate analysis questions from multiple views; and 3) BI technical skills, examining the models' use of technical knowledge to address real-world data analysis challenges. BIBench comprises 11 sub-tasks, spanning three categories of task types: classification, extraction, and generation. Additionally, we've developed BIChat, a domain-specific dataset with over a million data points, to fine-tune LLMs. We will release BIBenchmark, BIChat, and the evaluation scripts at \url{//github.com/cubenlp/BIBench}. This benchmark aims to provide a measure for in-depth analysis of LLM abilities and foster the advancement of LLMs in the field of data analysis.
Even though virtual testing of Autonomous Vehicles (AVs) has been well recognized as essential for safety assessment, AV simulators are still undergoing active development. One particularly challenging question is to effectively include the Sensing and Perception (S&P) subsystem into the simulation loop. In this article, we define Perception Error Models (PEM), a virtual simulation component that can enable the analysis of the impact of perception errors on AV safety, without the need to model the sensors themselves. We propose a generalized data-driven procedure towards parametric modeling and evaluate it using Apollo, an open-source driving software, and nuScenes, a public AV dataset. Additionally, we implement PEMs in SVL, an open-source vehicle simulator. Furthermore, we demonstrate the usefulness of PEM-based virtual tests, by evaluating camera, LiDAR, and camera-LiDAR setups. Our virtual tests highlight limitations in the current evaluation metrics, and the proposed approach can help study the impact of perception errors on AV safety.
Generative Pre-trained Transformer (GPT) models have exhibited exciting progress in their capabilities, capturing the interest of practitioners and the public alike. Yet, while the literature on the trustworthiness of GPT models remains limited, practitioners have proposed employing capable GPT models for sensitive applications such as healthcare and finance -- where mistakes can be costly. To this end, this work proposes a comprehensive trustworthiness evaluation for large language models with a focus on GPT-4 and GPT-3.5, considering diverse perspectives -- including toxicity, stereotype bias, adversarial robustness, out-of-distribution robustness, robustness on adversarial demonstrations, privacy, machine ethics, and fairness. Based on our evaluations, we discover previously unpublished vulnerabilities to trustworthiness threats. For instance, we find that GPT models can be easily misled to generate toxic and biased outputs and leak private information in both training data and conversation history. We also find that although GPT-4 is usually more trustworthy than GPT-3.5 on standard benchmarks, GPT-4 is more vulnerable given jailbreaking system or user prompts, potentially because GPT-4 follows (misleading) instructions more precisely. Our work illustrates a comprehensive trustworthiness evaluation of GPT models and sheds light on the trustworthiness gaps. Our benchmark is publicly available at //decodingtrust.github.io/ ; our dataset can be previewed at //huggingface.co/datasets/AI-Secure/DecodingTrust ; a concise version of this work is at //openreview.net/pdf?id=kaHpo8OZw2 .
Large Language Models (LLMs) have quickly risen to prominence due to their ability to perform at or close to the state-of-the-art in a variety of fields while handling natural language. An important field of research is the application of such models at the cybersecurity context. This survey aims to identify where in the field of cybersecurity LLMs have already been applied, the ways in which they are being used and their limitations in the field. Finally, suggestions are made on how to improve such limitations and what can be expected from these systems once these limitations are overcome.
Quality Assurance (QA) aims to prevent mistakes and defects in manufactured products and avoid problems when delivering products or services to customers. QA for AI systems, however, poses particular challenges, given their data-driven and non-deterministic nature as well as more complex architectures and algorithms. While there is growing empirical evidence about practices of machine learning in industrial contexts, little is known about the challenges and best practices of quality assurance for AI systems (QA4AI). In this paper, we report on a mixed-method study of QA4AI in industry practice from various countries and companies. Through interviews with fifteen industry practitioners and a validation survey with 50 practitioner responses, we studied the concerns as well as challenges and best practices in ensuring the QA4AI properties reported in the literature, such as correctness, fairness, interpretability and others. Our findings suggest correctness as the most important property, followed by model relevance, efficiency and deployability. In contrast, transferability (applying knowledge learned in one task to another task), security and fairness are not paid much attention by practitioners compared to other properties. Challenges and solutions are identified for each QA4AI property. For example, interviewees highlighted the trade-off challenge among latency, cost and accuracy for efficiency (latency and cost are parts of efficiency concern). Solutions like model compression are proposed. We identified 21 QA4AI practices across each stage of AI development, with 10 practices being well recognized and another 8 practices being marginally agreed by the survey practitioners.
Recent advances in instruction-tuned Large Vision-Language Models (LVLMs) have imbued the models with the ability to generate high-level, image-grounded explanations with ease. While such capability is largely attributed to the rich world knowledge contained within the Large Language Models (LLMs), our work reveals their shortcomings in fine-grained visual categorization (FGVC) across six different benchmark settings. Most recent state-of-the-art LVLMs like LLaVa-1.5, InstructBLIP and GPT-4V not only severely deteriorate in terms of classification performance, e.g., average drop of 65.58 in EM for Stanford Dogs for LLaVA-1.5, but also struggle to generate an accurate explanation with detailed attributes based on the concept that appears within an input image despite their capability to generate holistic image-level descriptions. In-depth analyses show that instruction-tuned LVLMs exhibit modality gap, showing discrepancy when given textual and visual inputs that correspond to the same concept, preventing the image modality from leveraging the rich parametric knowledge within the LLMs. In an effort to further the community's endeavor in this direction, we propose a multiple granularity attribute-centric evaluation benchmark, Finer, which aims to establish a ground to evaluate LVLMs' fine-grained visual comprehension ability and provide significantly improved explainability.
Recent embedding-based methods have achieved great successes in exploiting entity alignment from knowledge graph (KG) embeddings of multiple modalities. In this paper, we study embedding-based entity alignment (EEA) from a perspective of generative models. We show that EEA shares similarities with typical generative models and prove the effectiveness of the recently developed generative adversarial network (GAN)-based EEA methods theoretically. We then reveal that their incomplete objective limits the capacity on both entity alignment and entity synthesis (i.e., generating new entities). We mitigate this problem by introducing a generative EEA (GEEA) framework with the proposed mutual variational autoencoder (M-VAE) as the generative model. M-VAE enables entity conversion between KGs and generation of new entities from random noise vectors. We demonstrate the power of GEEA with theoretical analysis and empirical experiments on both entity alignment and entity synthesis tasks.
In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work.
An effective and efficient architecture performance evaluation scheme is essential for the success of Neural Architecture Search (NAS). To save computational cost, most of existing NAS algorithms often train and evaluate intermediate neural architectures on a small proxy dataset with limited training epochs. But it is difficult to expect an accurate performance estimation of an architecture in such a coarse evaluation way. This paper advocates a new neural architecture evaluation scheme, which aims to determine which architecture would perform better instead of accurately predict the absolute architecture performance. Therefore, we propose a \textbf{relativistic} architecture performance predictor in NAS (ReNAS). We encode neural architectures into feature tensors, and further refining the representations with the predictor. The proposed relativistic performance predictor can be deployed in discrete searching methods to search for the desired architectures without additional evaluation. Experimental results on NAS-Bench-101 dataset suggests that, sampling 424 ($0.1\%$ of the entire search space) neural architectures and their corresponding validation performance is already enough for learning an accurate architecture performance predictor. The accuracies of our searched neural architectures on NAS-Bench-101 and NAS-Bench-201 datasets are higher than that of the state-of-the-art methods and show the priority of the proposed method.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191