On Large Language Models' Hallucination with Regard to Known Facts

Large language models are successful in answering factoid questions but are also prone to hallucination.We investigate the phenomenon of LLMs possessing correct answer knowledge yet still hallucinating from the perspective of inference dynamics, an area not previously covered in studies on hallucinations.We are able to conduct this analysis via two key ideas.First, we identify the factual questions that query the same triplet knowledge but result in different answers. The difference between the model behaviors on the correct and incorrect outputs hence suggests the patterns when hallucinations happen. Second, to measure the pattern, we utilize mappings from the residual streams to vocabulary space. We reveal the different dynamics of the output token probabilities along the depths of layers between the correct and hallucinated cases. In hallucinated cases, the output token's information rarely demonstrates abrupt increases and consistent superiority in the later stages of the model. Leveraging the dynamic curve as a feature, we build a classifier capable of accurately detecting hallucinatory predictions with an 88\% success rate. Our study shed light on understanding the reasons for LLMs' hallucinations on their known facts, and more importantly, on accurately predicting when they are hallucinating.

MANGO: A Benchmark for Evaluating Mapping and Navigation Abilities of Large Language Models

Large language models such as ChatGPT and GPT-4 have recently achieved astonishing performance on a variety of natural language processing tasks. In this paper, we propose MANGO, a benchmark to evaluate their capabilities to perform text-based mapping and navigation. Our benchmark includes 53 mazes taken from a suite of textgames: each maze is paired with a walkthrough that visits every location but does not cover all possible paths. The task is question-answering: for each maze, a large language model reads the walkthrough and answers hundreds of mapping and navigation questions such as "How should you go to Attic from West of House?" and "Where are we if we go north and east from Cellar?". Although these questions are easy to humans, it turns out that even GPT-4, the best-to-date language model, performs poorly at answering them. Further, our experiments suggest that a strong mapping and navigation ability would benefit large language models in performing relevant downstream tasks, such as playing textgames. Our MANGO benchmark will facilitate future research on methods that improve the mapping and navigation capabilities of language models. We host our leaderboard, data, code, and evaluation program at https://mango.ttic.edu and https://github.com/oaklight/mango/.

Unsupervised Information Refinement Training of Large Language Models for Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating additional information from retrieval. However, studies have shown that LLMs still face challenges in effectively using the retrieved information, even ignoring it or being misled by it. The key reason is that the training of LLMs does not clearly make LLMs learn how to utilize input retrieved texts with varied quality. In this paper, we propose a novel perspective that considers the role of LLMs in RAG as ``Information Refiner'', which means that regardless of correctness, completeness, or usefulness of retrieved texts, LLMs can consistently integrate knowledge within the retrieved texts and model parameters to generate the texts that are more concise, accurate, and complete than the retrieved texts. To this end, we propose an information refinement training method named InFO-RAG that optimizes LLMs for RAG in an unsupervised manner. InFO-RAG is low-cost and general across various tasks. Extensive experiments on zero-shot prediction of 11 datasets in diverse tasks including Question Answering, Slot-Filling, Language Modeling, Dialogue, and Code Generation show that InFO-RAG improves the performance of LLaMA2 by an average of 9.39\% relative points. InFO-RAG also shows advantages in in-context learning and robustness of RAG.

Graph Representation of Narrative Context: Coherence Dependency via Retrospective Questions

This work introduces a novel and practical paradigm for narrative comprehension, stemming from the observation that individual passages within narratives are often cohesively related than being isolated. We therefore propose to formulate a graph upon narratives dubbed NARCO that depicts a task-agnostic coherence dependency of the entire context. Especially, edges in NARCO encompass retrospective free-form questions between two context snippets reflecting high-level coherent relations, inspired by the cognitive perception of humans who constantly reinstate relevant events from prior context. Importantly, our graph is instantiated through our designed two-stage LLM prompting, thereby without reliance on human annotations. We present three unique studies on its practical utility, examining the edge efficacy via recap identification, local context augmentation via plot retrieval, and broader applications exemplified by long document QA. Experiments suggest that our approaches leveraging NARCO yield performance boost across all three tasks.

Identifying Factual Inconsistency in Summaries: Towards Effective Utilization of Large Language Model

Factual inconsistency poses a significant hurdle for the commercial deployment of abstractive summarizers. Under this Large Language Model (LLM) era, this work focuses around two important questions: what is the best way to leverage LLM for factual inconsistency detection, and how could we distill a smaller LLM with both high efficiency and efficacy? Three zero-shot paradigms are firstly proposed and evaluated across five diverse datasets: direct inference on the entire summary or each summary window; entity verification through question generation and answering. Experiments suggest that LLM itself is capable to resolve this task train-free under the proper paradigm design, surpassing strong trained baselines by 2.8% on average. To further promote practical utility, we then propose training strategies aimed at distilling smaller open-source LLM that learns to score the entire summary at once with high accuracy, which outperforms the zero-shot approaches by much larger LLM, serving as an effective and efficient ready-to-use scorer.

Previously on the Stories: Recap Snippet Identification for Story Reading

Similar to the "previously-on" scenes in TV shows, recaps can help book reading by recalling the readers' memory about the important elements in previous texts to better understand the ongoing plot. Despite its usefulness, this application has not been well studied in the NLP community. We propose the first benchmark on this useful task called Recap Snippet Identification with a hand-crafted evaluation dataset. Our experiments show that the proposed task is challenging to PLMs, LLMs, and proposed methods as the task requires a deep understanding of the plot correlation between snippets.

Plot Retrieval as an Assessment of Abstract Semantic Association

Retrieving relevant plots from the book for a query is a critical task, which can improve the reading experience and efficiency of readers. Readers usually only give an abstract and vague description as the query based on their own understanding, summaries, or speculations of the plot, which requires the retrieval model to have a strong ability to estimate the abstract semantic associations between the query and candidate plots. However, existing information retrieval (IR) datasets cannot reflect this ability well. In this paper, we propose Plot Retrieval, a labeled dataset to train and evaluate the performance of IR models on the novel task Plot Retrieval. Text pairs in Plot Retrieval have less word overlap and more abstract semantic association, which can reflect the ability of the IR models to estimate the abstract semantic association, rather than just traditional lexical or semantic matching. Extensive experiments across various lexical retrieval, sparse retrieval, dense retrieval, and cross-encoder methods compared with human studies on Plot Retrieval show current IR models still struggle in capturing abstract semantic association between texts. Plot Retrieval can be the benchmark for further research on the semantic association modeling ability of IR models.

Personality Understanding of Fictional Characters during Book Reading

Comprehending characters' personalities is a crucial aspect of story reading. As readers engage with a story, their understanding of a character evolves based on new events and information; and multiple fine-grained aspects of personalities can be perceived. This leads to a natural problem of situated and fine-grained personality understanding. The problem has not been studied in the NLP field, primarily due to the lack of appropriate datasets mimicking the process of book reading. We present the first labeled dataset PersoNet for this problem. Our novel annotation strategy involves annotating user notes from online reading apps as a proxy for the original books. Experiments and human studies indicate that our dataset construction is both efficient and accurate; and our task heavily relies on long-term context to achieve accurate predictions for both machines and humans. The dataset is available at https://github.com/Gorov/personet_acl23.

Towards Coherent Image Inpainting Using Denoising Diffusion Implicit Models

Image inpainting refers to the task of generating a complete, natural image based on a partially revealed reference image. Recently, many research interests have been focused on addressing this problem using fixed diffusion models. These approaches typically directly replace the revealed region of the intermediate or final generated images with that of the reference image or its variants. However, since the unrevealed regions are not directly modified to match the context, it results in incoherence between revealed and unrevealed regions. To address the incoherence problem, a small number of methods introduce a rigorous Bayesian framework, but they tend to introduce mismatches between the generated and the reference images due to the approximation errors in computing the posterior distributions. In this paper, we propose COPAINT, which can coherently inpaint the whole image without introducing mismatches. COPAINT also uses the Bayesian framework to jointly modify both revealed and unrevealed regions, but approximates the posterior distribution in a way that allows the errors to gradually drop to zero throughout the denoising steps, thus strongly penalizing any mismatches with the reference image. Our experiments verify that COPAINT can outperform the existing diffusion-based methods under both objective and subjective metrics. The codes are available at https://github.com/UCSB-NLP-Chang/CoPaint/.