Since language models produce fake text quickly and easily, there is an oversupply of such content in the public domain. The degree of sophistication and writing style has reached a point where differentiating between human authored and machine-generated content is nearly impossible. As a result, works generated by language models rather than human authors have gained significant media attention and stirred controversy.Concerns regarding the possible influence of advanced language models on society have also arisen, needing a fuller knowledge of these processes. Natural language generation (NLG) and generative pre-trained transformer (GPT) models have revolutionized a variety of sectors: the scope not only permeated throughout journalism and customer service but also reached academia. To mitigate the hazardous implications that may arise from the use of these models, preventative measures must be implemented, such as providing human agents with the capacity to distinguish between artificially made and human composed texts utilizing automated systems and possibly reverse-engineered language models. Furthermore, to ensure a balanced and responsible approach, it is critical to have a full grasp of the socio-technological ramifications of these breakthroughs. This literature survey aims to compile and synthesize accomplishments and developments in the aforementioned work, while also identifying future prospects. It also gives an overview of machine-generated text trends and explores the larger societal implications. Ultimately, this survey intends to contribute to the development of robust and effective approaches for resolving the issues connected with the usage and detection of machine-generated text by exploring the interplay between the capabilities of language models and their possible implications.
Recent advancements in large language models (LLMs) have significantly enhanced capabilities in natural language processing and artificial intelligence. These models, including GPT-3.5 and LLaMA-2, have revolutionized text generation, translation, and question-answering tasks due to the transformative Transformer model. Despite their widespread use, LLMs present challenges such as ethical dilemmas when models are compelled to respond inappropriately, susceptibility to phishing attacks, and privacy violations. This paper addresses these challenges by introducing a multi-pronged approach that includes: 1) filtering sensitive vocabulary from user input to prevent unethical responses; 2) detecting role-playing to halt interactions that could lead to 'prison break' scenarios; 3) implementing custom rule engines to restrict the generation of prohibited content; and 4) extending these methodologies to various LLM derivatives like Multi-Model Large Language Models (MLLMs). Our approach not only fortifies models against unethical manipulations and privacy breaches but also maintains their high performance across tasks. We demonstrate state-of-the-art performance under various attack prompts, without compromising the model's core functionalities. Furthermore, the introduction of differentiated security levels empowers users to control their personal data disclosure. Our methods contribute to reducing social risks and conflicts arising from technological abuse, enhance data protection, and promote social equity. Collectively, this research provides a framework for balancing the efficiency of question-answering systems with user privacy and ethical standards, ensuring a safer user experience and fostering trust in AI technology.
Data-to-text (D2T) generation is a crucial task in many natural language understanding (NLU) applications and forms the foundation of task-oriented dialog systems. In the context of conversational AI solutions that can work directly with local data on the user's device, architectures utilizing large pre-trained language models (PLMs) are impractical for on-device deployment due to a high memory footprint. To this end, we propose TrICy, a novel lightweight framework for an enhanced D2T task that generates text sequences based on the intent in context and may further be guided by user-provided triggers. We leverage an attention-copy mechanism to predict out-of-vocabulary (OOV) words accurately. Performance analyses on E2E NLG dataset (BLEU: 66.43%, ROUGE-L: 70.14%), WebNLG dataset (BLEU: Seen 64.08%, Unseen 52.35%), and our Custom dataset related to text messaging applications, showcase our architecture's effectiveness. Moreover, we show that by leveraging an optional trigger input, data-to-text generation quality increases significantly and achieves the new SOTA score of 69.29% BLEU for E2E NLG. Furthermore, our analyses show that TrICy achieves at least 24% and 3% improvement in BLEU and METEOR respectively over LLMs like GPT-3, ChatGPT, and Llama 2. We also demonstrate that in some scenarios, performance improvement due to triggers is observed even when they are absent in training.
Large language models (LLMs), especially when instruction-tuned for chat, have become part of our daily lives, freeing people from the process of searching, extracting, and integrating information from multiple sources by offering a straightforward answer to a variety of questions in a single place. Unfortunately, in many cases, LLM responses are factually incorrect, which limits their applicability in real-world scenarios. As a result, research on evaluating and improving the factuality of LLMs has attracted a lot of attention recently. In this survey, we critically analyze existing work with the aim to identify the major challenges and their associated causes, pointing out to potential solutions for improving the factuality of LLMs, and analyzing the obstacles to automated factuality evaluation for open-ended text generation. We further offer an outlook on where future research should go.
Distribution matching methods for language model alignment such as Generation with Distributional Control (GDC) and Distributional Policy Gradient (DPG) have not received the same level of attention in reinforcement learning from human feedback (RLHF) as contrastive methods such as Sequence Likelihood Calibration (SLiC), Direct Preference Optimization (DPO) and its variants. We identify high variance of the gradient estimate as the primary reason for the lack of success of these methods and propose a self-normalized baseline to reduce the variance. We further generalize the target distribution in DPG, GDC and DPO by using Bayes' rule to define the reward-conditioned posterior. The resulting approach, referred to as BRAIn - Bayesian Reward-conditioned Amortized Inference acts as a bridge between distribution matching methods and DPO and significantly outperforms prior art in summarization and Antropic HH tasks.
Large language models (LLMs) have shown remarkable advances in language generation and understanding but are also prone to exhibiting harmful social biases. While recognition of these behaviors has generated an abundance of bias mitigation techniques, most require modifications to the training data, model parameters, or decoding strategy, which may be infeasible without access to a trainable model. In this work, we leverage the zero-shot capabilities of LLMs to reduce stereotyping in a technique we introduce as zero-shot self-debiasing. With two approaches, self-debiasing via explanation and self-debiasing via reprompting, we show that self-debiasing can significantly reduce the degree of stereotyping across nine different social groups while relying only on the LLM itself and a simple prompt, with explanations correctly identifying invalid assumptions and reprompting delivering the greatest reductions in bias. We hope this work opens inquiry into other zero-shot techniques for bias mitigation.
This paper introduces a novel Bayesian learning model to explain the behavior of Large Language Models (LLMs), focusing on their core optimization metric of next token prediction. We develop a theoretical framework based on an ideal generative text model represented by a multinomial transition probability matrix with a prior, and examine how LLMs approximate this matrix. Key contributions include: (i) a continuity theorem relating embeddings to multinomial distributions, (ii) a demonstration that LLM text generation aligns with Bayesian learning principles, (iii) an explanation for the emergence of in-context learning in larger models, (iv) empirical validation using visualizations of next token probabilities from an instrumented Llama model Our findings provide new insights into LLM functioning, offering a statistical foundation for understanding their capabilities and limitations. This framework has implications for LLM design, training, and application, potentially guiding future developments in the field.
This research explores strategies for steering the output of large language models (LLMs) towards specific styles, such as sentiment, emotion, or writing style, by adding style vectors to the activations of hidden layers during text generation. We show that style vectors can be simply computed from recorded layer activations for input texts in a specific style in contrast to more complex training-based approaches. Through a series of experiments, we demonstrate the effectiveness of activation engineering using such style vectors to influence the style of generated text in a nuanced and parameterisable way, distinguishing it from prompt engineering. The presented research constitutes a significant step towards developing more adaptive and effective AI-empowered interactive systems.
Generating natural language text from graph-structured data is essential for conversational information seeking. Semantic triples derived from knowledge graphs can serve as a valuable source for grounding responses from conversational agents by providing a factual basis for the information they communicate. This is especially relevant in the context of large language models, which offer great potential for conversational interaction but are prone to hallucinating, omitting, or producing conflicting information. In this study, we conduct an empirical analysis of conversational large language models in generating natural language text from semantic triples. We compare four large language models of varying sizes with different prompting techniques. Through a series of benchmark experiments on the WebNLG dataset, we analyze the models' performance and identify the most common issues in the generated predictions. Our findings show that the capabilities of large language models in triple verbalization can be significantly improved through few-shot prompting, post-processing, and efficient fine-tuning techniques, particularly for smaller models that exhibit lower zero-shot performance.
Evaluating natural language generation (NLG) is a vital but challenging problem in natural language processing. Traditional evaluation metrics mainly capturing content (e.g. n-gram) overlap between system outputs and references are far from satisfactory, and large language models (LLMs) such as ChatGPT have demonstrated great potential in NLG evaluation in recent years. Various automatic evaluation methods based on LLMs have been proposed, including metrics derived from LLMs, prompting LLMs, fine-tuning LLMs, and human-LLM collaborative evaluation. In this survey, we first give a taxonomy of LLM-based NLG evaluation methods, and discuss their pros and cons, respectively. Lastly, we discuss several open problems in this area and point out future research directions.
In the field of NLP, Large Language Models (LLMs) have markedly enhanced performance across a variety of tasks. However, the comprehensive evaluation of LLMs remains an inevitable challenge for the community. Recently, the adoption of Multiple Choice Question Answering (MCQA) as a benchmark for assessing LLMs has gained considerable traction. However, concerns regarding the robustness of this evaluative method persist. Building upon previous discussions on the issue of \textit{variability}, we reveal an additional dimension of concern: LLMs may perform MCQA by selecting the least incorrect option rather than distinctly correct. This observation suggests that LLMs might regard multiple options as correct, which could undermine the reliability of MCQA as a metric for evaluating LLMs. To address this challenge, we introduce an enhanced dataset augmentation method for MCQA, termed MCQA+, to provide a more accurate reflection of the model performance, thereby highlighting the necessity for more sophisticated evaluation mechanisms in the assessment of LLM capabilities.
Feature attribution methods (FAs), such as gradients and attention, are widely employed approaches to derive the importance of all input features to the model predictions. Existing work in natural language processing has mostly focused on developing and testing FAs for encoder-only language models (LMs) in classification tasks. However, it is unknown if it is faithful to use these FAs for decoder-only models on text generation, due to the inherent differences between model architectures and task settings respectively. Moreover, previous work has demonstrated that there is no `one-wins-all' FA across models and tasks. This makes the selection of a FA computationally expensive for large LMs since input importance derivation often requires multiple forward and backward passes including gradient computations that might be prohibitive even with access to large compute. To address these issues, we present a model-agnostic FA for generative LMs called Recursive Attribution Generator (ReAGent). Our method updates the token importance distribution in a recursive manner. For each update, we compute the difference in the probability distribution over the vocabulary for predicting the next token between using the original input and using a modified version where a part of the input is replaced with RoBERTa predictions. Our intuition is that replacing an important token in the context should have resulted in a larger change in the model's confidence in predicting the token than replacing an unimportant token. Our method can be universally applied to any generative LM without accessing internal model weights or additional training and fine-tuning, as most other FAs require. We extensively compare the faithfulness of ReAGent with seven popular FAs across six decoder-only LMs of various sizes. The results show that our method consistently provides more faithful token importance distributions.
Quantifying uncertainty in automatically generated text is important for letting humans check potential hallucinations and making systems more reliable. Conformal prediction is an attractive framework to provide predictions imbued with statistical guarantees, however, its application to text generation is challenging since any i.i.d. assumptions are not realistic. In this paper, we bridge this gap by leveraging recent results on non-exchangeable conformal prediction, which still ensures bounds on coverage. The result, non-exchangeable conformal nucleus sampling, is a novel extension of the conformal prediction framework to generation based on nearest neighbors. Our method can be used post-hoc for an arbitrary model without extra training and supplies token-level, calibrated prediction sets equipped with statistical guarantees. Experiments in machine translation and language modeling show encouraging results in generation quality. By also producing tighter prediction sets with good coverage, we thus give a more theoretically principled way to perform sampling with conformal guarantees.
Large language models (LLMs) have exhibited remarkable capabilities in text generation tasks. However, the utilization of these models carries inherent risks, including but not limited to plagiarism, the dissemination of fake news, and issues in educational exercises. Although several detectors have been proposed to address these concerns, their effectiveness against adversarial perturbations, specifically in the context of student essay writing, remains largely unexplored. This paper aims to bridge this gap by constructing AIG-ASAP, an AI-generated student essay dataset, employing a range of text perturbation methods that are expected to generate high-quality essays while evading detection. Through empirical experiments, we assess the performance of current AIGC detectors on the AIG-ASAP dataset. The results reveal that the existing detectors can be easily circumvented using straightforward automatic adversarial attacks. Specifically, we explore word substitution and sentence substitution perturbation methods that effectively evade detection while maintaining the quality of the generated essays. This highlights the urgent need for more accurate and robust methods to detect AI-generated student essays in the education domain.
In this work, we introduce the Multiple Embedding Model for EHR (MEME), an approach that serializes multimodal EHR tabular data into text using pseudo-notes, mimicking clinical text generation. This conversion not only preserves better representations of categorical data and learns contexts but also enables the effective employment of pretrained foundation models for rich feature representation. To address potential issues with context length, our framework encodes embeddings for each EHR modality separately. We demonstrate the effectiveness of MEME by applying it to several decision support tasks within the Emergency Department across multiple hospital systems. Our findings indicate that MEME outperforms traditional machine learning, EHR-specific foundation models, and general LLMs, highlighting its potential as a general and extendible EHR representation strategy.
The rise of Large Language Models (LLMs) has affected various disciplines that got beyond mere text generation. Going beyond their textual nature, this project proposal aims to investigate the interaction between LLMs and non-verbal communication, specifically focusing on gestures. The proposal sets out a plan to examine the proficiency of LLMs in deciphering both explicit and implicit non-verbal cues within textual prompts and their ability to associate these gestures with various contextual factors. The research proposes to test established psycholinguistic study designs to construct a comprehensive dataset that pairs textual prompts with detailed gesture descriptions, encompassing diverse regional variations, and semantic labels. To assess LLMs' comprehension of gestures, experiments are planned, evaluating their ability to simulate human behaviour in order to replicate psycholinguistic experiments. These experiments consider cultural dimensions and measure the agreement between LLM-identified gestures and the dataset, shedding light on the models' contextual interpretation of non-verbal cues (e.g. gestures).
State-space models are a low-complexity alternative to transformers for encoding long sequences and capturing long-term dependencies. We propose LOCOST: an encoder-decoder architecture based on state-space models for conditional text generation with long context inputs. With a computational complexity of $O(L \log L)$, this architecture can handle significantly longer sequences than state-of-the-art models that are based on sparse attention patterns. We evaluate our model on a series of long document abstractive summarization tasks. The model reaches a performance level that is 93-96% comparable to the top-performing sparse transformers of the same size while saving up to 50% memory during training and up to 87% during inference. Additionally, LOCOST effectively handles input texts exceeding 600K tokens at inference time, setting new state-of-the-art results on full-book summarization and opening new perspectives for long input processing.
We present "Paramanu", a family of novel language models (LM) for Indian languages, consisting of auto-regressive monolingual, bilingual, and multilingual models pretrained from scratch. Currently, it covers 10 languages (Assamese, Bangla, Hindi, Konkani, Maithili, Marathi, Odia, Sanskrit, Tamil, Telugu) across 5 scripts (Bangla, Devanagari, Odia, Tamil, Telugu). The models are pretrained on a single GPU with context size of 1024 and vary in size from 13.29 million (M) to 367.5 M parameters. We proposed a RoPE embedding scaling method that enables us to pretrain language models from scratch at larger sequence length context size than typical GPU memory permits. We also introduced a novel efficient Indic tokenizer, "mBharat", using a combination of BPE and Unigram, achieving the least fertility score and the ability to tokenize unseen languages in both the same script & Roman script. We also proposed and performed language-specific tokenization for multilingual models & domain-specific tokenization for monolingual models. To address the "curse of multilinguality" in our mParamanu model, we pretrained on comparable corpora based on typological grouping within the same script. Our findings show a language transfer phenomenon from low-resource to high-resource languages within languages of the same script & typology. Human evaluations for open-ended text generation demonstrated that Paramanu models outperformed several LLMs, despite being 20 to 64 times smaller. We created instruction-tuning datasets & instruction-tuned our models on 23,000 instructions in respective languages. Comparisons with multilingual LLMs across various benchmarks for natural language (NL) understanding, NL inference, & reading comprehension highlight the advantages of our models; leads to the conclusion that high quality generative LM are possible without high amount of compute power & enormous number of parameters.
In this report, we explore the potential for text diffusion to replace autoregressive (AR) decoding for the training and deployment of large language models (LLMs). We are particularly interested to see whether pretrained AR models can be transformed into text diffusion models through a lightweight adaptation procedure we call ``AR2Diff''. We begin by establishing a strong baseline setup for training text diffusion models. Comparing across multiple architectures and pretraining objectives, we find that training a decoder-only model with a prefix LM objective is best or near-best across several tasks. Building on this finding, we test various transfer learning setups for text diffusion models. On machine translation, we find that text diffusion underperforms the standard AR approach. However, on code synthesis and extractive QA, we find diffusion models trained from scratch outperform AR models in many cases. We also observe quality gains from AR2Diff -- adapting AR models to use diffusion decoding. These results are promising given that text diffusion is relatively underexplored and can be significantly faster than AR decoding for long text generation.
Instruction-tuned Large Language Models (LLMs) have recently showcased remarkable advancements in their ability to generate fitting responses to natural language instructions. However, many current works rely on manual evaluation to judge the quality of generated responses. Since such manual evaluation is time-consuming, it does not easily scale to the evaluation of multiple models and model variants. In this short paper, we propose a straightforward but remarkably effective evaluation metric called SemScore, in which we directly compare model outputs to gold target responses using semantic textual similarity (STS). We conduct a comparative evaluation of the model outputs of 12 prominent instruction-tuned LLMs using 8 widely-used evaluation metrics for text generation. We find that our proposed SemScore metric outperforms all other, in many cases more complex, evaluation metrics in terms of correlation to human evaluation. These findings indicate the utility of our proposed metric for the evaluation of instruction-tuned LLMs.