Recently, the instruction-tuning of large language models is a crucial area of research in the field of natural language processing. Due to resource and cost limitations, several researchers have employed parameter-efficient tuning techniques, such as LoRA, for instruction tuning, and have obtained encouraging results In comparison to full-parameter fine-tuning, LoRA-based tuning demonstrates salient benefits in terms of training costs. In this study, we undertook experimental comparisons between full-parameter fine-tuning and LoRA-based tuning methods, utilizing LLaMA as the base model. The experimental results show that the selection of the foundational model, training dataset scale, learnable parameter quantity, and model training cost are all important factors. We hope that the experimental conclusions of this paper can provide inspiration for training large language models, especially in the field of Chinese, and help researchers find a better trade-off strategy between training cost and model performance. To facilitate the reproduction of the paper's results, the dataset, model and code will be released.
Online news platforms often use personalized news recommendation methods to help users discover articles that align with their interests. These methods typically predict a matching score between a user and a candidate article to reflect the user's preference for the article. Some previous works have used language model techniques, such as the attention mechanism, to capture users' interests based on their past behaviors, and to understand the content of articles. However, these existing model architectures require adjustments if additional information is taken into account. Pre-trained large language models, which can better capture word relationships and comprehend contexts, have seen a significant development in recent years, and these pre-trained models have the advantages of transfer learning and reducing the training time for downstream tasks. Meanwhile, prompt learning is a newly developed technique that leverages pre-trained language models by building task-specific guidance for output generations. To leverage textual information in news articles, this paper introduces the pre-trained large language model and prompt-learning to the community of news recommendation. The proposed model "prompt-based news recommendation" (PBNR) treats the personalized news recommendation as a text-to-text language task and designs personalized prompts to adapt to the pre-trained language model -- text-to-text transfer transformer (T5). Experimental studies using the Microsoft News dataset show that PBNR is capable of making accurate recommendations by taking into account various lengths of past behaviors of different users. PBNR can also easily adapt to new information without changing the model architecture and the training objective. Additionally, PBNR can make recommendations based on users' specific requirements, allowing human-computer interaction in the news recommendation field.
Supervised approaches for learning spatio-temporal scene graphs (STSG) from video are greatly hindered due to their reliance on STSG-annotated videos, which are labor-intensive to construct at scale. Is it feasible to instead use readily available video captions as weak supervision? To address this question, we propose LASER, a neuro-symbolic framework to enable training STSG generators using only video captions. LASER employs large language models to first extract logical specifications with rich spatio-temporal semantic information from video captions. LASER then trains the underlying STSG generator to align the predicted STSG with the specification. The alignment algorithm overcomes the challenges of weak supervision by leveraging a differentiable symbolic reasoner and using a combination of contrastive, temporal, and semantics losses. The overall approach efficiently trains low-level perception models to extract a fine-grained STSG that conforms to the video caption. In doing so, it enables a novel methodology for learning STSGs without tedious annotations. We evaluate our method on three video datasets: OpenPVSG, 20BN, and MUGEN. Our approach demonstrates substantial improvements over fully-supervised baselines, achieving a unary predicate prediction accuracy of 27.78% (+12.65%) and a binary recall@5 of 0.42 (+0.22) on OpenPVSG. Additionally, LASER exceeds baselines by 7% on 20BN and 5.2% on MUGEN in terms of overall predicate prediction accuracy.
This contribution analyzes the self-perception and political biases of OpenAI's Large Language Model ChatGPT. Taking into account the first small-scale reports and studies that have emerged, claiming that ChatGPT is politically biased towards progressive and libertarian points of view, this contribution aims to provide further clarity on this subject. For this purpose, ChatGPT was asked to answer the questions posed by the political compass test as well as similar questionnaires that are specific to the respective politics of the G7 member states. These eight tests were repeated ten times each and revealed that ChatGPT seems to hold a bias towards progressive views. The political compass test revealed a bias towards progressive and libertarian views, with the average coordinates on the political compass being (-6.48, -5.99) (with (0, 0) the center of the compass, i.e., centrism and the axes ranging from -10 to 10), supporting the claims of prior research. The political questionnaires for the G7 member states indicated a bias towards progressive views but no significant bias between authoritarian and libertarian views, contradicting the findings of prior reports, with the average coordinates being (-3.27, 0.58). In addition, ChatGPT's Big Five personality traits were tested using the OCEAN test and its personality type was queried using the Myers-Briggs Type Indicator (MBTI) test. Finally, the maliciousness of ChatGPT was evaluated using the Dark Factor test. These three tests were also repeated ten times each, revealing that ChatGPT perceives itself as highly open and agreeable, has the Myers-Briggs personality type ENFJ, and is among the 15% of test-takers with the least pronounced dark traits.
We describe two systems currently being developed that use large language models for the automatized correction of (i) exercises in translating back and forth between natural language and the languages of propositional logic and first-order predicate logic and (ii) exercises in writing simple arguments in natural language in non-mathematical scenarios.
Individual human decision-makers may benefit from different forms of support to improve decision outcomes, but when each form of support will yield better outcomes? In this work, we posit that personalizing access to decision support tools can be an effective mechanism for instantiating the appropriate use of AI assistance. Specifically, we propose the general problem of learning a decision support policy that, for a given input, chooses which form of support to provide to decision-makers for whom we initially have no prior information. We develop $\texttt{Modiste}$, an interactive tool to learn personalized decision support policies. $\texttt{Modiste}$ leverages stochastic contextual bandit techniques to personalize a decision support policy for each decision-maker and supports extensions to the multi-objective setting to account for auxiliary objectives like the cost of support. We find that personalized policies outperform offline policies, and, in the cost-aware setting, reduce the incurred cost with minimal degradation to performance. Our experiments include various realistic forms of support (e.g., expert consensus and predictions from a large language model) on vision and language tasks. Our human subject experiments validate our computational experiments, demonstrating that personalization can yield benefits in practice for real users, who interact with $\texttt{Modiste}$.
This paper assesses the accuracy, reliability and bias of the Large Language Model (LLM) ChatGPT-4 on the text analysis task of classifying the political affiliation of a Twitter poster based on the content of a tweet. The LLM is compared to manual annotation by both expert classifiers and crowd workers, generally considered the gold standard for such tasks. We use Twitter messages from United States politicians during the 2020 election, providing a ground truth against which to measure accuracy. The paper finds that ChatGPT-4 has achieves higher accuracy, higher reliability, and equal or lower bias than the human classifiers. The LLM is able to correctly annotate messages that require reasoning on the basis of contextual knowledge, and inferences around the author's intentions - traditionally seen as uniquely human abilities. These findings suggest that LLM will have substantial impact on the use of textual data in the social sciences, by enabling interpretive research at a scale.
Large foundation language models have shown their versatility in being able to be adapted to perform a wide variety of downstream tasks, such as text generation, sentiment analysis, semantic search etc. However, training such large foundational models is a non-trivial exercise that requires a significant amount of compute power and expertise from machine learning and systems experts. As models get larger, these demands are only increasing. Sparsity is a promising technique to relieve the compute requirements for training. However, sparsity introduces new challenges in training the sparse model to the same quality as the dense counterparts. Furthermore, sparsity drops the operation intensity and introduces irregular memory access patterns that makes it challenging to efficiently utilize compute resources. This paper demonstrates an end-to-end training flow on a large language model - 13 billion GPT - using sparsity and dataflow. The dataflow execution model and architecture enables efficient on-chip irregular memory accesses as well as native kernel fusion and pipelined parallelism that helps recover device utilization. We show that we can successfully train GPT 13B to the same quality as the dense GPT 13B model, while achieving an end-end speedup of 4.5x over dense A100 baseline.
We demonstrate the potential of the state-of-the-art OpenAI GPT-4 large language model to engage in meaningful interactions with Astronomy papers using in-context prompting. To optimize for efficiency, we employ a distillation technique that effectively reduces the size of the original input paper by 50\%, while maintaining the paragraph structure and overall semantic integrity. We then explore the model's responses using a multi-document context (ten distilled documents). Our findings indicate that GPT-4 excels in the multi-document domain, providing detailed answers contextualized within the framework of related research findings. Our results showcase the potential of large language models for the astronomical community, offering a promising avenue for further exploration, particularly the possibility of utilizing the models for hypothesis generation.
This study evaluates the potential of a large language model for aiding in generation of semantic feature norms - a critical tool for evaluating conceptual structure in cognitive science. Building from an existing human-generated dataset, we show that machine-verified norms capture aspects of conceptual structure beyond what is expressed in human norms alone, and better explain human judgments of semantic similarity amongst items that are distally related. The results suggest that LLMs can greatly enhance traditional methods of semantic feature norm verification, with implications for our understanding of conceptual representation in humans and machines.
Reinforcement Learning from Human Feedback (RLHF) facilitates the alignment of large language models with human preferences, significantly enhancing the quality of interactions between humans and models. InstructGPT implements RLHF through several stages, including Supervised Fine-Tuning (SFT), reward model training, and Proximal Policy Optimization (PPO). However, PPO is sensitive to hyperparameters and requires multiple models in its standard implementation, making it hard to train and scale up to larger parameter counts. In contrast, we propose a novel learning paradigm called RRHF, which scores sampled responses from different sources via a logarithm of conditional probabilities and learns to align these probabilities with human preferences through ranking loss. RRHF can leverage sampled responses from various sources including the model responses from itself, other large language model responses, and human expert responses to learn to rank them. RRHF only needs 1 to 2 models during tuning and can efficiently align language models with human preferences robustly without complex hyperparameter tuning. Additionally, RRHF can be considered an extension of SFT and reward model training while being simpler than PPO in terms of coding, model counts, and hyperparameters. We evaluate RRHF on the Helpful and Harmless dataset, demonstrating comparable alignment performance with PPO by reward model score and human labeling. Extensive experiments show that the performance of RRHF is highly related to sampling quality which suggests RRHF is a best-of-n learner. Codes available at https://github.com/GanjinZero/RRHF.
Large language models (LLMs) have achieved impressive performance on code generation. However, for complex programming tasks, generating the correct solution in one go becomes challenging, thus some prior works have designed program repair approaches to improve code generation performance. In this work, we propose Self-Debugging, which teaches a large language model to debug its predicted program via few-shot demonstrations. In particular, we demonstrate that Self-Debugging can teach the large language model to perform rubber duck debugging; i.e., without any human feedback on the code correctness or error messages, the model is able to identify its mistakes by investigating the execution results and explaining the generated code in natural language. Self-Debugging achieves the state-of-the-art performance on several code generation benchmarks, including the Spider dataset for text-to-SQL generation, TransCoder for C++-to-Python translation, and MBPP for text-to-Python generation. On the Spider benchmark where there are no unit tests to verify the correctness of predictions, Self-Debugging with code explanation consistently improves the baseline by 2-3%, and improves the prediction accuracy on problems of the hardest level by 9%. On TransCoder and MBPP where unit tests are available, Self-Debugging improves the baseline accuracy by up to 12%. Meanwhile, by leveraging feedback messages and reusing failed predictions, Self-Debugging notably improves sample efficiency, and can match or outperform baseline models that generate more than 10x candidate programs.
We propose LLMA, an LLM accelerator to losslessly speed up Large Language Model (LLM) inference with references. LLMA is motivated by the observation that there are abundant identical text spans between the decoding result by an LLM and the reference that is available in many real world scenarios (e.g., retrieved documents). LLMA first selects a text span from the reference and copies its tokens to the decoder and then efficiently checks the tokens' appropriateness as the decoding result in parallel within one decoding step. The improved computational parallelism allows LLMA to achieve over 2x speed-up for LLMs with identical generation results as greedy decoding in many practical generation scenarios where significant overlap between in-context reference and outputs exists (e.g., search engines and multi-turn conversations).
The widespread usage of computer-based assessments and individualized learning platforms has resulted in an increased demand for the rapid production of high-quality items. Automated item generation (AIG), the process of using item models to generate new items with the help of computer technology, was proposed to reduce reliance on human subject experts at each step of the process. AIG has been used in test development for some time. Still, the use of machine learning algorithms has introduced the potential to improve the efficiency and effectiveness of the process greatly. The approach presented in this paper utilizes OpenAI's latest transformer-based language model, GPT-3, to generate reading passages. Existing reading passages were used in carefully engineered prompts to ensure the AI-generated text has similar content and structure to a fourth-grade reading passage. For each prompt, we generated multiple passages, the final passage was selected according to the Lexile score agreement with the original passage. In the final round, the selected passage went through a simple revision by a human editor to ensure the text was free of any grammatical and factual errors. All AI-generated passages, along with original passages were evaluated by human judges according to their coherence, appropriateness to fourth graders, and readability.
Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents--computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent's experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors: for example, starting with only a single user-specified notion that one agent wants to throw a Valentine's Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture--observation, planning, and reflection--each contribute critically to the believability of agent behavior. By fusing large language models with computational, interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior.
The ChatGPT, a lite and conversational variant of Generative Pretrained Transformer 4 (GPT-4) developed by OpenAI, is one of the milestone Large Language Models (LLMs) with billions of parameters. LLMs have stirred up much interest among researchers and practitioners in their impressive skills in natural language processing tasks, which profoundly impact various fields. This paper mainly discusses the future applications of LLMs in dentistry. We introduce two primary LLM deployment methods in dentistry, including automated dental diagnosis and cross-modal dental diagnosis, and examine their potential applications. Especially, equipped with a cross-modal encoder, a single LLM can manage multi-source data and conduct advanced natural language reasoning to perform complex clinical operations. We also present cases to demonstrate the potential of a fully automatic Multi-Modal LLM AI system for dentistry clinical application. While LLMs offer significant potential benefits, the challenges, such as data privacy, data quality, and model bias, need further study. Overall, LLMs have the potential to revolutionize dental diagnosis and treatment, which indicates a promising avenue for clinical application and research in dentistry.
In recent years, single cell RNA sequencing has become a widely used technique to study cellular diversity and function. However, accurately annotating cell types from single cell data has been a challenging task, as it requires extensive knowledge of cell biology and gene function. The emergence of large language models such as ChatGPT and New Bing in 2023 has revolutionized this process by integrating the scientific literature and providing accurate annotations of cell types. This breakthrough enables researchers to conduct literature reviews more efficiently and accurately, and can potentially uncover new insights into cell type annotation. By using ChatGPT to annotate single cell data, we can relate rare cell type to their function and reveal specific differentiation trajectories of cell subtypes that were previously overlooked. This can have important applications in understanding cancer progression, mammalian development, and stem cell differentiation, and can potentially lead to the discovery of key cells that interrupt the differentiation pathway and solve key problems in the life sciences. Overall, the future of cell type annotation in single cell data looks promising and the Large Language model will be an important milestone in the history of single cell analysis.
Heatmaps are widely used to interpret deep neural networks, particularly for computer vision tasks, and the heatmap-based explainable AI (XAI) techniques are a well-researched topic. However, most studies concentrate on enhancing the quality of the generated heatmap or discovering alternate heatmap generation techniques, and little effort has been devoted to making heatmap-based XAI automatic, interactive, scalable, and accessible. To address this gap, we propose a framework that includes two modules: (1) context modelling and (2) reasoning. We proposed a template-based image captioning approach for context modelling to create text-based contextual information from the heatmap and input data. The reasoning module leverages a large language model to provide explanations in combination with specialised knowledge. Our qualitative experiments demonstrate the effectiveness of our framework and heatmap captioning approach. The code for the proposed template-based heatmap captioning approach will be publicly available.
The main objective of this paper is to identify the major research areas of ChatGPT through term and keyword co-occurrence network mapping techniques. For conducting the present study, total of 577 publications were retrieved from the Lens database for the network visualization. The findings of the study showed that chatgpt occurrence in maximum number of times followed by its related terms such as artificial intelligence, large language model, gpt, study etc. This study will be helpful to library and information science as well as computer or information technology professionals.
For a natural language problem that requires some non-trivial reasoning to solve, there are at least two ways to do it using a large language model (LLM). One is to ask it to solve it directly. The other is to use it to extract the facts from the problem text and then use a theorem prover to solve it. In this note, we compare the two methods using ChatGPT and GPT4 on a series of logic word puzzles, and conclude that the latter is the right approach.