In this work, we present TextHarmony, a unified and versatile multimodal generative model proficient in comprehending and generating visual text. Simultaneously generating images and texts typically results in performance degradation due to the inherent inconsistency between vision and language modalities. To overcome this challenge, existing approaches resort to modality-specific data for supervised fine-tuning, necessitating distinct model instances. We propose Slide-LoRA, which dynamically aggregates modality-specific and modality-agnostic LoRA experts, partially decoupling the multimodal generation space. Slide-LoRA harmonizes the generation of vision and language within a singular model instance, thereby facilitating a more unified generative process. Additionally, we develop a high-quality image caption dataset, DetailedTextCaps-100K, synthesized with a sophisticated closed-source MLLM to enhance visual text generation capabilities further. Comprehensive experiments across various benchmarks demonstrate the effectiveness of the proposed approach. Empowered by Slide-LoRA, TextHarmony achieves comparable performance to modality-specific fine-tuning results with only a 2% increase in parameters and shows an average improvement of 2.5% in visual text comprehension tasks and 4.0% in visual text generation tasks. Our work delineates the viability of an integrated approach to multimodal generation within the visual text domain, setting a foundation for subsequent inquiries. Code is available at https://github.com/bytedance/TextHarmony.
Legal Large Language Models (LLMs) have shown promise in providing legal consultations to non-experts. However, most existing Chinese legal consultation models are based on single-agent systems, which differ from real-world legal consultations, where multiple professionals collaborate to offer more tailored responses. To better simulate real consultations, we propose LawLuo, a multi-agent framework for multi-turn Chinese legal consultations. LawLuo includes four agents: the receptionist agent, which assesses user intent and selects a lawyer agent; the lawyer agent, which interacts with the user; the secretary agent, which organizes conversation records and generates consultation reports; and the boss agent, which evaluates the performance of the lawyer and secretary agents to ensure optimal results. These agents' interactions mimic the operations of real law firms. To train them to follow different legal instructions, we developed distinct fine-tuning datasets. We also introduce a case graph-based RAG to help the lawyer agent address vague user inputs. Experimental results show that LawLuo outperforms baselines in generating more personalized and professional responses, handling ambiguous queries, and following legal instructions in multi-turn conversations. Our full code and constructed datasets will be open-sourced upon paper acceptance.
Using questions in written text is an effective strategy to enhance readability. However, what makes an active reading question good, what the linguistic role of these questions is, and what is their impact on human reading remains understudied. We introduce GuidingQ, a dataset of 10K in-text questions from textbooks and scientific articles. By analyzing the dataset, we present a comprehensive understanding of the use, distribution, and linguistic characteristics of these questions. Then, we explore various approaches to generate such questions using language models. Our results highlight the importance of capturing inter-question relationships and the challenge of question position identification in generating these questions. Finally, we conduct a human study to understand the implication of such questions on reading comprehension. We find that the generated questions are of high quality and are almost as effective as human-written questions in terms of improving readers' memorization and comprehension.
Extractive question answering over clinical text is a crucial need to help deal with the deluge of clinical text generated in hospitals. While encoder models (e.g., BERT) have been popular for this reading comprehension task, recently encoder-decoder models (e.g., T5) are on the rise. There is also the emergence of preference optimization techniques to align decoder-only LLMs with human preferences. In this paper, we combine encoder-decoder models with the direct preference optimization (DPO) method to improve over prior state of the art for the RadQA radiology question answering task by 12-15 F1 points. To the best of our knowledge, this effort is the first to show that DPO method also works for reading comprehension via novel heuristics to generate preference data without human inputs.
Structured data, rich in logical and relational information, has the potential to enhance the reasoning abilities of large language models (LLMs). Still, its integration poses a challenge due to the risk of overwhelming LLMs with excessive tokens and irrelevant context information. To address this, we propose Struct-X, a novel framework that operates through five key phases: ``read-model-fill-reflect-reason'' efficiently enabling LLMs to utilize structured data. It begins by encoding structured data into a topological space using graph embeddings, followed by filling in missing entity information with knowledge retrieval modules, and filtering out irrelevant tokens via a self-supervised module. The final phase involves constructing a topological network with selected tokens to further reduce the total token length for more effective LLM inference. Additionally, Struct-X includes an Auxiliary Module trained to generate prompts, aiding LLMs in analyzing structured data. Extensive experiments on benchmarks, including the knowledge graph question-answer task and the long document reading comprehension task, show that Struct-X notably improves LLM reasoning, demonstrating the effectiveness of structured data augmentation in improving LLM inference with complex input context.
Large Language Models (LLMs) are able to provide assistance on a wide range of information-seeking tasks. However, model outputs may be misleading, whether unintentionally or in cases of intentional deception. We investigate the ability of LLMs to be deceptive in the context of providing assistance on a reading comprehension task, using LLMs as proxies for human users. We compare outcomes of (1) when the model is prompted to provide truthful assistance, (2) when it is prompted to be subtly misleading, and (3) when it is prompted to argue for an incorrect answer. Our experiments show that GPT-4 can effectively mislead both GPT-3.5-Turbo and GPT-4, with deceptive assistants resulting in up to a 23% drop in accuracy on the task compared to when a truthful assistant is used. We also find that providing the user model with additional context from the passage partially mitigates the influence of the deceptive model. This work highlights the ability of LLMs to produce misleading information and the effects this may have in real-world situations.
Recently, there has been a growing interest among large language model (LLM) developers in LLM-based document reading systems, which enable users to upload their own documents and pose questions related to the document contents, going beyond simple reading comprehension tasks. Consequently, these systems have been carefully designed to tackle challenges such as file parsing, metadata extraction, multi-modal information understanding and long-context reading. However, no current benchmark exists to evaluate their performance in such scenarios, where a raw file and questions are provided as input, and a corresponding response is expected as output. In this paper, we introduce DocBench, a new benchmark designed to evaluate LLM-based document reading systems. Our benchmark involves a meticulously crafted process, including the recruitment of human annotators and the generation of synthetic questions. It includes 229 real documents and 1,102 questions, spanning across five different domains and four major types of questions. We evaluate both proprietary LLM-based systems accessible via web interfaces or APIs, and a parse-then-read pipeline employing open-source LLMs. Our evaluations reveal noticeable gaps between existing LLM-based document reading systems and human performance, underscoring the challenges of developing proficient systems. To summarize, DocBench aims to establish a standardized benchmark for evaluating LLM-based document reading systems under diverse real-world scenarios, thereby guiding future advancements in this research area.
Despite the remarkable success of LLMs in English, there is a significant gap in performance in non-English languages. In order to address this, we introduce a novel recipe for creating a multilingual synthetic instruction tuning dataset, sPhinX, which is created by selectively translating instruction response pairs from English into 50 languages. We test the effectiveness of sPhinx by using it to fine-tune two state-of-the-art models, Mistral-7B and Phi-Small and then evaluating them across a comprehensive suite of multilingual benchmarks that test reasoning, question answering, reading comprehension and machine translation. Our results show that Mistral-7B and Phi-Small fine-tuned with sPhinX perform better on an average by 5%pt for both the models when compared to the base variants of these models. We also devise a strategy to incorporate N-shot examples in each fine-tuning sample which further boosts the performance of these models by 9%pt and 4%pt respectively respectively compared to vanilla fine-tuning. To show efficacy of our data curation approach, we also directly translate our original dataset to the target languages, and observe an increase of 7%pt and 4%pt on both the models respectively. sPhinX outperforms other multilingual instruction tuning datasets in both efficiency and diversity, reducing dataset creation costs. It also maintains strong performance on standard English LLM benchmarks, with minimal regression.
The emergence of large language models (LLMs) is a milestone in generative artificial intelligence, achieving significant success in text comprehension and generation tasks. Despite the tremendous success of LLMs in many downstream tasks, they suffer from severe hallucination problems, posing significant challenges to the practical applications of LLMs. Most of the works about LLMs' hallucinations focus on data quality. Self-attention is a core module in transformer-based LLMs, while its potential relationship with LLMs' hallucination has been hardly investigated. To fill this gap, we study this problem from a causal perspective. We propose a method to intervene in LLMs' self-attention layers and maintain their structures and sizes intact. Specifically, we disable different self-attention layers in several popular open-source LLMs and then compare their degrees of hallucination with the original ones. We evaluate the intervened LLMs on hallucination assessment benchmarks and conclude that disabling some specific self-attention layers in the front or tail of the LLMs can alleviate hallucination issues. The study paves a new way for understanding and mitigating LLMs' hallucinations.
Recent advancements in AI, particularly in large language models (LLMs) like ChatGPT, Claude, and Gemini, have prompted questions about their proximity to Artificial General Intelligence (AGI). This study compares LLM performance on educational benchmarks with Americans' average educational attainment and literacy levels, using data from the U.S. Census Bureau and technical reports. Results show that LLMs significantly outperform human benchmarks in tasks such as undergraduate knowledge and advanced reading comprehension, indicating substantial progress toward AGI. However, true AGI requires broader cognitive assessments. The study highlights the implications for AI development, education, and societal impact, emphasizing the need for ongoing research and ethical considerations.
Synthetic data is becoming increasingly important for accelerating the development of language models, both large and small. Despite several successful use cases, researchers also raised concerns around model collapse and drawbacks of imitating other models. This discrepancy can be attributed to the fact that synthetic data varies in quality and diversity. Effective use of synthetic data usually requires significant human effort in curating the data. We focus on using synthetic data for post-training, specifically creating data by powerful models to teach a new skill or behavior to another model, we refer to this setting as Generative Teaching. We introduce AgentInstruct, an extensible agentic framework for automatically creating large amounts of diverse and high-quality synthetic data. AgentInstruct can create both the prompts and responses, using only raw data sources like text documents and code files as seeds. We demonstrate the utility of AgentInstruct by creating a post training dataset of 25M pairs to teach language models different skills, such as text editing, creative writing, tool usage, coding, reading comprehension, etc. The dataset can be used for instruction tuning of any base model. We post-train Mistral-7b with the data. When comparing the resulting model Orca-3 to Mistral-7b-Instruct (which uses the same base model), we observe significant improvements across many benchmarks. For example, 40% improvement on AGIEval, 19% improvement on MMLU, 54% improvement on GSM8K, 38% improvement on BBH and 45% improvement on AlpacaEval. Additionally, it consistently outperforms other models such as LLAMA-8B-instruct and GPT-3.5-turbo.
This paper introduces MalAlgoQA, a novel dataset designed to evaluate the counterfactual reasoning capabilities of Large Language Models (LLMs) through a pedagogical approach. The dataset comprises mathematics and reading comprehension questions, each accompanied by four answer choices and their corresponding rationales. At the heart of MalAlgoQA are ``malgorithms'' - rationales behind incorrect answer choices that represent flawed yet logically coherent reasoning paths. These malgorithms serve as counterfactual scenarios, allowing us to assess an LLM's ability to identify and analyze flawed reasoning patterns. We propose the Malgorithm Identification task, where LLMs are assessed based on their ability to identify corresponding malgorithm given an incorrect answer choice. To evaluate the model performance, we introduce two metrics: Algorithm Identification Accuracy (AIA) for correct answer rationale identification, and Malgorithm Identification Accuracy (MIA) for incorrect answer rationale identification. Our experiments reveal that state-of-the-art LLMs exhibit significant performance drops in MIA compared to AIA, highlighting the challenges in counterfactual reasoning. Surprisingly, we find that the chain-of-thought prompting technique not only fails to consistently enhance MIA but can sometimes lead to underperformance compared to simple prompting. These findings have important implications for developing LLMs with improved counterfactual reasoning, particularly relevant for AI-powered tutoring systems, where identifying and addressing student misconceptions is essential. MalAlgoQA dataset is available \href{https://github.com/luffycodes/MalAlgoQA-Dataset}{here}.
With an increasing social demand for fine-grained sentiment analysis (SA), implicit sentiment analysis (ISA) poses a significant challenge with the absence of salient cue words in expressions. It necessitates reliable reasoning to understand how the sentiment is aroused and thus determine implicit sentiments. In the era of Large Language Models (LLMs), Encoder-Decoder (ED) LLMs have gained popularity to serve as backbone models for SA applications, considering impressive text comprehension and reasoning ability among diverse tasks. On the other hand, Decoder-only (DO) LLMs exhibit superior natural language generation and in-context learning capabilities. However, their responses may contain misleading or inaccurate information. To identify implicit sentiment with reliable reasoning, this study proposes RVISA, a two-stage reasoning framework that harnesses the generation ability of DO LLMs and the reasoning ability of ED LLMs to train an enhanced reasoner. Specifically, we adopt three-hop reasoning prompting to explicitly furnish sentiment elements as cues. The generated rationales are utilized to fine-tune an ED LLM into a skilled reasoner. Additionally, we develop a straightforward yet effective verification mechanism to ensure the reliability of the reasoning learning. We evaluated the proposed method on two benchmark datasets and achieved state-of-the-art results in ISA performance.
Instruction-tuned Large Language Models (LLMs) have achieved remarkable performance across various benchmark tasks. While providing instructions to LLMs for guiding their generations is user-friendly, assessing their instruction-following capabilities is still unclarified due to a lack of evaluation metrics. In this paper, we focus on evaluating the instruction-following ability of LLMs in the context of story-ending generation, which requires diverse and context-specific instructions. We propose an automatic evaluation pipeline that utilizes a machine reading comprehension (MRC) model to determine whether the generated story-ending reflects instruction. Our findings demonstrate that our proposed metric aligns with human evaluation. Furthermore, our experiments confirm that recent open-source LLMs can achieve instruction-following performance close to GPT-3.5, as assessed through automatic evaluation.
Natural language processing has seen rapid progress over the past decade. Due to the speed of developments, some practices get established without proper evaluation. Considering one such case and focusing on reading comprehension, we ask our first research question: 1) How does the order of inputs -- i.e., question and context -- affect model performance? Additionally, given recent advancements in input emphasis, we ask a second research question: 2) Does emphasizing either the question, the context, or both enhance performance? Experimenting with 9 large language models across 3 datasets, we find that presenting the context before the question improves model performance, with an accuracy increase of up to $31\%$. Furthermore, emphasizing the context yields superior results compared to question emphasis, and in general, emphasizing parts of the input is particularly effective for addressing questions that models lack the parametric knowledge to answer. Experimenting with both prompt-based and attention-based emphasis methods, we additionally find that the best method is surprisingly simple: it only requires concatenating a few tokens to the input and results in an accuracy improvement of up to $36\%$, allowing smaller models to outperform their significantly larger counterparts.
Large language models (LLMs) excel in generating coherent text, but they often struggle with context awareness, leading to inaccuracies in tasks requiring faithful adherence to provided information. We introduce FastMem, a novel method designed to enhance instruction fine-tuned LLMs' context awareness through fast memorization of the prompt. FastMem maximizes the likelihood of the prompt before inference by updating only the last Feed-Forward Network (FFN) module. This targeted approach ensures efficient optimization without overfitting, significantly improving the model's ability to comprehend and accurately follow the context. Our experiments demonstrate substantial gains in reading comprehension, text summarization and adherence to output structures. For instance, FastMem improves the accuracy of Llama 3-8B-Inst on the NQ-SWAP dataset from 59.1% to 71.6%, and reduces the output structure failure rate of Qwen 1.5-4B-Chat from 34.9% to 25.5%. Extensive experimental results highlight FastMem's potential to offer a robust solution to enhance the reliability and accuracy of LLMs in various applications. Our code is available at: https://github.com/IAAR-Shanghai/FastMem
Large Language Models (LLMs) have recently demonstrated remarkable performance in various Natural Language Processing (NLP) applications, such as sentiment analysis, content generation, and personalized recommendations. Despite their impressive capabilities, there remains a significant need for systematic studies concerning the practical application of LLMs in industrial settings, as well as the specific requirements and challenges related to their deployment in these contexts. This need is particularly critical for Machine Reading Comprehension (MCR), where factual, concise, and accurate responses are required. To date, most MCR rely on Small Language Models (SLMs) or Recurrent Neural Networks (RNNs) such as Long Short-Term Memory (LSTM). This trend is evident in the SQuAD2.0 rankings on the Papers with Code table. This article presents a comparative analysis between open-source LLMs and proprietary models on this task, aiming to identify light and open-source alternatives that offer comparable performance to proprietary models.
Commonsense reasoning is fundamentally based on multimodal knowledge. However, existing large language models (LLMs) are primarily trained using textual data only, limiting their ability to incorporate essential visual information. In contrast, Visual Language Models, which excel at visually-oriented tasks, often fail at non-visual tasks such as basic commonsense reasoning. This divergence highlights a critical challenge - the integration of robust visual understanding with foundational text-based language reasoning. To this end, we introduce a method aimed at enhancing LLMs' visual commonsense. Specifically, our method generates multiple images based on the input text prompt and integrates these into the model's decision-making process by mixing their prediction probabilities. To facilitate multimodal grounded language modeling, we employ a late-fusion layer that combines the projected visual features with the output of a pre-trained LLM conditioned on text only. This late-fusion layer enables predictions based on comprehensive image-text knowledge as well as text only when this is required. We evaluate our approach using several visual commonsense reasoning tasks together with traditional NLP tasks, including common sense reasoning and reading comprehension. Our experimental results demonstrate significant superiority over existing baselines. When applied to recent state-of-the-art LLMs (e.g., Llama3), we observe improvements not only in visual common sense but also in traditional NLP benchmarks. Code and models are available under https://github.com/guyyariv/vLMIG.
Despite their vast capabilities, Large Language Models (LLMs) often struggle with generating reliable outputs, frequently producing high-confidence inaccuracies known as hallucinations. Addressing this challenge, our research introduces InternalInspector, a novel framework designed to enhance confidence estimation in LLMs by leveraging contrastive learning on internal states including attention states, feed-forward states, and activation states of all layers. Unlike existing methods that primarily focus on the final activation state, InternalInspector conducts a comprehensive analysis across all internal states of every layer to accurately identify both correct and incorrect prediction processes. By benchmarking InternalInspector against existing confidence estimation methods across various natural language understanding and generation tasks, including factual question answering, commonsense reasoning, and reading comprehension, InternalInspector achieves significantly higher accuracy in aligning the estimated confidence scores with the correctness of the LLM's predictions and lower calibration error. Furthermore, InternalInspector excels at HaluEval, a hallucination detection benchmark, outperforming other internal-based confidence estimation methods in this task.
Large Language Models (LLMs), already shown to ace various text comprehension tasks, have also remarkably been shown to tackle table comprehension tasks without specific training. Building on earlier studies of LLMs for tabular tasks, we probe how in-context learning (ICL), model scale, instruction tuning, and domain bias affect Tabular QA (TQA) robustness by testing LLMs, under diverse augmentations and perturbations, on diverse domains: Wikipedia-based $\textbf{WTQ}$, financial $\textbf{TAT-QA}$, and scientific $\textbf{SCITAB}$. Although instruction tuning and larger, newer LLMs deliver stronger, more robust TQA performance, data contamination and reliability issues, especially on $\textbf{WTQ}$, remain unresolved. Through an in-depth attention analysis, we reveal a strong correlation between perturbation-induced shifts in attention dispersion and the drops in performance, with sensitivity peaking in the model's middle layers. We highlight the need for improved interpretable methodologies to develop more reliable LLMs for table comprehension.