Previous approaches to persona simulation large language models (LLMs) have typically relied on learning basic biographical information, or using limited role-play dialogue datasets to capture a character's responses. However, a holistic representation of an individual goes beyond surface-level facts or conversations to deeper thoughts and thinking. In this work, we introduce CharacterBot, a model designed to replicate both the linguistic patterns and distinctive thought processes of a character. Using Lu Xun, a renowned Chinese writer, as a case study, we propose four training tasks derived from his 17 essay collections. These include a pre-training task focused on mastering external linguistic structures and knowledge, as well as three fine-tuning tasks: multiple-choice question answering, generative question answering, and style transfer, each aligning the LLM with Lu Xun's internal ideation and writing style. To optimize learning across these tasks, we introduce a CharLoRA parameter updating mechanism, where a general linguistic style expert collaborates with other task-specific experts to better study both the language style and the understanding of deeper thoughts. We evaluate CharacterBot on three tasks for linguistic accuracy and opinion comprehension, demonstrating that it significantly outperforms the baselines on our adapted metrics. We hope that this work inspires future research on deep character persona simulation LLM.
When addressing professional questions in the biomedical domain, humans typically acquire multiple pieces of information as evidence and engage in multifaceted evidence analysis to provide high-quality answers. Current LLM-based answer generation methods lack a detailed definition and learning process for evidence analysis, leading to the risk of error propagation and hallucinations while using evidence. Although increasing the parameter size of LLMs can alleviate these issues, it also presents challenges in model training and deployment with limited resources. In this study, we propose EvidenceMap, which aims to enable a tiny pre-trained language model to explicitly learn multiple aspects of biomedical evidence, including supportive evaluation, logical correlation and content summarization, thereby latently guiding a small generative model (around 3B parameters) to provide textual responses. Experimental results demonstrate that our method, fine-tuning a language model with 66M parameters, exceeds the RAG method with an 8B LLM by 19.9% and 5.7% in reference-based quality and accuracy, respectively.
We develop a principled procedure for determining when a large language model (LLM) should abstain from responding (e.g., by saying "I don't know") in a general domain, instead of resorting to possibly "hallucinating" a non-sensical or incorrect answer. Building on earlier approaches that use self-consistency as a more reliable measure of model confidence, we propose using the LLM itself to self-evaluate the similarity between each of its sampled responses for a given query. We then further leverage conformal prediction techniques to develop an abstention procedure that benefits from rigorous theoretical guarantees on the hallucination rate (error rate). Experimentally, our resulting conformal abstention method reliably bounds the hallucination rate on various closed-book, open-domain generative question answering datasets, while also maintaining a significantly less conservative abstention rate on a dataset with long responses (Temporal Sequences) compared to baselines using log-probability scores to quantify uncertainty, while achieveing comparable performance on a dataset with short answers (TriviaQA). To evaluate the experiments automatically, one needs to determine if two responses are equivalent given a question. Following standard practice, we use a thresholded similarity function to determine if two responses match, but also provide a method for calibrating the threshold based on conformal prediction, with theoretical guarantees on the accuracy of the match prediction, which might be of independent interest.
To address the hallucination in generative question answering (GQA) where the answer can not be derived from the document, we propose a novel evidence-enhanced triplet generation framework, EATQA, encouraging the model to predict all the combinations of (Question, Evidence, Answer) triplet by flipping the source pair and the target label to understand their logical relationships, i.e., predict Answer(A), Question(Q), and Evidence(E) given a QE, EA, and QA pairs, respectively. Furthermore, we bridge the distribution gap to distill the knowledge from evidence in inference stage. Our framework ensures the model to learn the logical relation between query, evidence and answer, which simultaneously improves the evidence generation and query answering. In this paper, we apply EATQA to LLama and it outperforms other LLMs-based methods and hallucination mitigation approaches on two challenging GQA benchmarks. Further analysis shows that our method not only keeps prior knowledge within LLM, but also mitigates hallucination and generates faithful answers.
Large language models (LLMs) have demonstrated impressive performance in mathematical and commonsense reasoning tasks using chain-of-thought (CoT) prompting techniques. But can they perform emotional reasoning by concatenating `Let's think step-by-step' to the input prompt? In this paper we investigate this question along with introducing a novel approach to zero-shot emotion detection and emotional reasoning using LLMs. Existing state of the art zero-shot approaches rely on textual entailment models to choose the most appropriate emotion label for an input text. We argue that this strongly restricts the model to a fixed set of labels which may not be suitable or sufficient for many applications where emotion analysis is required. Instead, we propose framing the problem of emotion analysis as a generative question-answering (QA) task. Our approach uses a two step methodology of generating relevant context or background knowledge to answer the emotion detection question step-by-step. Our paper is the first work on using a generative approach to jointly address the tasks of emotion detection and emotional reasoning for texts. We evaluate our approach on two popular emotion detection datasets and also release the fine-grained emotion labels and explanations for further training and fine-tuning of emotional reasoning systems.
Reducing the `$\textit{hallucination}$' problem of Large Language Models (LLMs) is crucial for their wide applications. A comprehensive and fine-grained measurement of the hallucination is the first key step for the governance of this issue but is under-explored in the community. Thus, we present $\textbf{ANAH}$, a bilingual dataset that offers $\textbf{AN}$alytical $\textbf{A}$nnotation of $\textbf{H}$allucinations in LLMs within Generative Question Answering. Each answer sentence in our dataset undergoes rigorous annotation, involving the retrieval of a reference fragment, the judgment of the hallucination type, and the correction of hallucinated content. ANAH consists of ~12k sentence-level annotations for ~4.3k LLM responses covering over 700 topics, constructed by a human-in-the-loop pipeline. Thanks to the fine granularity of the hallucination annotations, we can quantitatively confirm that the hallucinations of LLMs progressively accumulate in the answer and use ANAH to train and evaluate hallucination annotators. We conduct extensive experiments on studying generative and discriminative annotators and show that, although current open-source LLMs have difficulties in fine-grained hallucination annotation, the generative annotator trained with ANAH can surpass all open-source LLMs and GPT-3.5, obtain performance competitive with GPT-4, and exhibits better generalization ability on unseen questions.
With the rapid advancement of natural language processing technologies, generative artificial intelligence techniques, represented by large language models (LLMs), are gaining increasing prominence and demonstrating significant potential for applications in safety engineering. However, fundamental LLMs face constraints such as limited training data coverage and unreliable responses. This study develops a vector database from 117 explosion accident reports in China spanning 2013 to 2023, employing techniques such as corpus segmenting and vector embedding. By utilizing the vector database, which outperforms the relational database in information retrieval quality, we provide LLMs with richer, more relevant knowledge. Comparative analysis of LLMs demonstrates that ChatSOS significantly enhances reliability, accuracy, and comprehensiveness, improves adaptability and clarification of responses. These results illustrate the effectiveness of supplementing LLMs with an external database, highlighting their potential to handle professional queries in safety engineering and laying a foundation for broader applications.
BACKGROUND: Radiology reports are typically written in a free-text format, making clinical information difficult to extract and use. Recently the adoption of structured reporting (SR) has been recommended by various medical societies thanks to the advantages it offers, e.g. standardization, completeness and information retrieval. We propose a pipeline to extract information from free-text radiology reports, that fits with the items of the reference SR registry proposed by a national society of interventional and medical radiology, focusing on CT staging of patients with lymphoma. METHODS: Our work aims to leverage the potential of Natural Language Processing (NLP) and Transformer-based models to deal with automatic SR registry filling. With the availability of 174 radiology reports, we investigate a rule-free generative Question Answering approach based on a domain-specific version of T5 (IT5). Two strategies (batch-truncation and ex-post combination) are implemented to comply with the model's context length limitations. Performance is evaluated in terms of strict accuracy, F1, and format accuracy, and compared with the widely used GPT-3.5 Large Language Model. A 5-point Likert scale questionnaire is used to collect human-expert feedback on the similarity between medical annotations and generated answers. RESULTS: The combination of fine-tuning and batch splitting allows IT5 to achieve notable results; it performs on par with GPT-3.5 albeit its size being a thousand times smaller in terms of parameters. Human-based assessment scores show a high correlation (Spearman's correlation coefficients>0.88, p-values<0.001) with AI performance metrics (F1) and confirm the superior ability of LLMs (i.e., GPT-3.5, 175B of parameters) in generating plausible human-like statements.
In this paper, we present the current progress of the project Verif.ai, an open-source scientific generative question-answering system with referenced and verified answers. The components of the system are (1) an information retrieval system combining semantic and lexical search techniques over scientific papers (PubMed), (2) a fine-tuned generative model (Mistral 7B) taking top answers and generating answers with references to the papers from which the claim was derived, and (3) a verification engine that cross-checks the generated claim and the abstract or paper from which the claim was derived, verifying whether there may have been any hallucinations in generating the claim. We are reinforcing the generative model by providing the abstract in context, but in addition, an independent set of methods and models are verifying the answer and checking for hallucinations. Therefore, we believe that by using our method, we can make scientists more productive, while building trust in the use of generative language models in scientific environments, where hallucinations and misinformation cannot be tolerated.
Temporal knowledge graph question answering (TKGQA) poses a significant challenge task, due to the temporal constraints hidden in questions and the answers sought from dynamic structured knowledge. Although large language models (LLMs) have made considerable progress in their reasoning ability over structured data, their application to the TKGQA task is a relatively unexplored area. This paper first proposes a novel generative temporal knowledge graph question answering framework, GenTKGQA, which guides LLMs to answer temporal questions through two phases: Subgraph Retrieval and Answer Generation. First, we exploit LLM's intrinsic knowledge to mine temporal constraints and structural links in the questions without extra training, thus narrowing down the subgraph search space in both temporal and structural dimensions. Next, we design virtual knowledge indicators to fuse the graph neural network signals of the subgraph and the text representations of the LLM in a non-shallow way, which helps the open-source LLM deeply understand the temporal order and structural dependencies among the retrieved facts through instruction tuning. Experimental results on two widely used datasets demonstrate the superiority of our model.
The present study introduces the knowledge-augmented generator, which is specifically designed to produce information that remains grounded in contextual knowledge, regardless of alterations in the context. Previous research has predominantly focused on examining hallucinations stemming from static input, such as in the domains of summarization or machine translation. However, our investigation delves into the faithfulness of generative question answering in the presence of dynamic knowledge. Our objective is to explore the existence of hallucinations arising from parametric memory when contextual knowledge undergoes changes, while also analyzing the underlying causes for their occurrence. In order to efficiently address this issue, we propose a straightforward yet effective measure for detecting such hallucinations. Intriguingly, our investigation uncovers that all models exhibit a tendency to generate previous answers as hallucinations. To gain deeper insights into the underlying causes of this phenomenon, we conduct a series of experiments that verify the critical role played by context in hallucination, both during training and testing, from various perspectives.
In many legal processes being able to action on the concrete implication of a legal question can be valuable to automating human review or signalling certain conditions (e.g., alerts around automatic renewal). To support such tasks, we present a form of legal question answering that seeks to return one (or more) fixed answers for a question about a contract clause. After showing that unstructured generative question answering can have questionable outcomes for such a task, we discuss our exploration methodology for legal question answering prompts using OpenAI's \textit{GPT-3.5-Turbo} and provide a summary of insights. Using insights gleaned from our qualitative experiences, we compare our proposed template prompts against a common semantic matching approach and find that our prompt templates are far more accurate despite being less reliable in the exact response return. With some additional tweaks to prompts and the use of in-context learning, we are able to further improve the performance of our proposed strategy while maximizing the reliability of responses as best we can.
This paper presents a simple and cost-effective method for synthesizing data to train question-answering systems. For training, fine-tuning GPT models is a common practice in resource-rich languages like English, however, it becomes challenging for non-English languages due to the scarcity of sufficient question-answer (QA) pairs. Existing approaches use question and answer generators trained on human-authored QA pairs, which involves substantial human expenses. In contrast, we use an instruct-tuned model to generate QA pairs in a zero-shot or few-shot manner. We conduct experiments to compare various strategies for obtaining QA pairs from the instruct-tuned model. The results demonstrate that a model trained on our proposed synthetic data achieves comparable performance to a model trained on manually curated datasets, without incurring human costs.
Current large language models (LLMs) can exhibit near-human levels of performance on many natural language-based tasks, including open-domain question answering. Unfortunately, at this time, they also convincingly hallucinate incorrect answers, so that responses to questions must be verified against external sources before they can be accepted at face value. In this paper, we report two simple experiments to automatically validate generated answers against a corpus. We base our experiments on questions and passages from the MS MARCO (V1) test collection, and a retrieval pipeline consisting of sparse retrieval, dense retrieval and neural rerankers. In the first experiment, we validate the generated answer in its entirety. After presenting a question to an LLM and receiving a generated answer, we query the corpus with the combination of the question + generated answer. We then present the LLM with the combination of the question + generated answer + retrieved answer, prompting it to indicate if the generated answer can be supported by the retrieved answer. In the second experiment, we consider the generated answer at a more granular level, prompting the LLM to extract a list of factual statements from the answer and verifying each statement separately. We query the corpus with each factual statement and then present the LLM with the statement and the corresponding retrieved evidence. The LLM is prompted to indicate if the statement can be supported and make necessary edits using the retrieved material. With an accuracy of over 80%, we find that an LLM is capable of verifying its generated answer when a corpus of supporting material is provided. However, manual assessment of a random sample of questions reveals that incorrect generated answers are missed by this verification process. While this verification process can reduce hallucinations, it can not entirely eliminate them.
In recent years, significant advancements in pre-trained language models have driven the creation of numerous non-English language variants, with a particular emphasis on encoder-only and decoder-only architectures. While Spanish language models based on BERT and GPT have demonstrated proficiency in natural language understanding and generation, there remains a noticeable scarcity of encoder-decoder models explicitly designed for sequence-to-sequence tasks, which aim to map input sequences to generate output sequences conditionally. This paper breaks new ground by introducing the implementation and evaluation of renowned encoder-decoder architectures exclusively pre-trained on Spanish corpora. Specifically, we present Spanish versions of BART, T5, and BERT2BERT-style models and subject them to a comprehensive assessment across various sequence-to-sequence tasks, including summarization, question answering, split-and-rephrase, dialogue, and translation. Our findings underscore the competitive performance of all models, with the BART- and T5-based models emerging as top performers across all tasks. We have made all models publicly available to the research community to foster future explorations and advancements in Spanish NLP: https://github.com/vgaraujov/Seq2Seq-Spanish-PLMs.
Pir\'a is a reading comprehension dataset focused on the ocean, the Brazilian coast, and climate change, built from a collection of scientific abstracts and reports on these topics. This dataset represents a versatile language resource, particularly useful for testing the ability of current machine learning models to acquire expert scientific knowledge. Despite its potential, a detailed set of baselines has not yet been developed for Pir\'a. By creating these baselines, researchers can more easily utilize Pir\'a as a resource for testing machine learning models across a wide range of question answering tasks. In this paper, we define six benchmarks over the Pir\'a dataset, covering closed generative question answering, machine reading comprehension, information retrieval, open question answering, answer triggering, and multiple choice question answering. As part of this effort, we have also produced a curated version of the original dataset, where we fixed a number of grammar issues, repetitions, and other shortcomings. Furthermore, the dataset has been extended in several new directions, so as to face the aforementioned benchmarks: translation of supporting texts from English into Portuguese, classification labels for answerability, automatic paraphrases of questions and answers, and multiple choice candidates. The results described in this paper provide several points of reference for researchers interested in exploring the challenges provided by the Pir\'a dataset.
We propose a framework - Prompt, Generate, Train (PGT) - to efficiently develop a generative question-answering model for open-book question-answering over a proprietary collection of text documents. The framework adapts a retriever augmented generation (RAG) model to the target domain using supervised fine-tuning and reinforcement learning with synthetic feedback in a few-shot setting. This, we hypothesize, will yield an aligned, uncertainty calibrated model that is competitive with GPT-4 based in-context retrieval augmented generation in generating relevant answers at lower serving costs. The framework's synthetic generation pipeline will generate synthetic training data comprising <passage, question, answer> tuples using an open-source LLM and a novel consistency filtering scheme. The pipeline will be designed to generate both abstractive and extractive questions that span the entire corpus. The framework proposes to fine-tune a smaller RAG model comprising a dense retriever (ColBERTv2) and a smaller sized LLM on the synthetic dataset. In parallel, the framework will train a Reward model to score domain grounded answers higher than hallucinated answers using an a priori relevance ordering of synthetically assembled samples. In the next phase, the framework will align the RAG model with the target domain using reinforcement learning (Proximal Policy Optimization). This step may improve the RAG model's ability to generate grounded answers and ignore out of domain questions. In the final phase, the framework will calibrate the model's uncertainty for extractive question-answers.
Large language models (LLMs) have garnered significant attention, but the definition of "large" lacks clarity. This paper focuses on medium-sized language models (MLMs), defined as having at least six billion parameters but less than 100 billion. The study evaluates MLMs regarding zero-shot generative question answering, which requires models to provide elaborate answers without external document retrieval. The paper introduces an own test dataset and presents results from human evaluation. Results show that combining the best answers from different MLMs yielded an overall correct answer rate of 82.7% which is better than the 60.9% of ChatGPT. The best MLM achieved 71.8% and has 33B parameters, which highlights the importance of using appropriate training data for fine-tuning rather than solely relying on the number of parameters. More fine-grained feedback should be used to further improve the quality of answers. The open source community is quickly closing the gap to the best commercial models.
Event argument extraction has long been studied as a sequential prediction problem with extractive-based methods, tackling each argument in isolation. Although recent work proposes generation-based methods to capture cross-argument dependency, they require generating and post-processing a complicated target sequence (template). Motivated by these observations and recent pretrained language models' capabilities of learning from demonstrations. We propose a retrieval-augmented generative QA model (R-GQA) for event argument extraction. It retrieves the most similar QA pair and augments it as prompt to the current example's context, then decodes the arguments as answers. Our approach outperforms substantially prior methods across various settings (i.e. fully supervised, domain transfer, and fewshot learning). Finally, we propose a clustering-based sampling strategy (JointEnc) and conduct a thorough analysis of how different strategies influence the few-shot learning performance. The implementations are available at https:// github.com/xinyadu/RGQA
Generative question answering (QA) models generate answers to questions either solely based on the parameters of the model (the closed-book setting) or additionally retrieving relevant evidence (the open-book setting). Generative QA models can answer some relatively complex questions, but the mechanism through which they do so is still poorly understood. We perform several studies aimed at better understanding the multi-hop reasoning capabilities of generative QA models. First, we decompose multi-hop questions into multiple corresponding single-hop questions, and find marked inconsistency in QA models' answers on these pairs of ostensibly identical question chains. Second, we find that models lack zero-shot multi-hop reasoning ability: when trained only on single-hop questions, models generalize poorly to multi-hop questions. Finally, we demonstrate that it is possible to improve models' zero-shot multi-hop reasoning capacity through two methods that approximate real multi-hop natural language (NL) questions by training on either concatenation of single-hop questions or logical forms (SPARQL). In sum, these results demonstrate that multi-hop reasoning does not emerge naturally in generative QA models, but can be encouraged by advances in training or modeling techniques.