We explore whether Large Language Models (LLMs) are capable of logical reasoning with distorted facts, which we call Deduction under Perturbed Evidence (DUPE). DUPE presents a unique challenge to LLMs since they typically rely on their parameters, which encode mostly accurate information, to reason and make inferences. However, in DUPE, LLMs must reason over manipulated or falsified evidence present in their prompts, which can result in false conclusions that are valid only under the manipulated evidence. Our goal with DUPE is to determine whether LLMs can arrive at these false conclusions and identify whether the dominant factor influencing the deduction process is the encoded data in the parameters or the manipulated evidence in the prompts. To evaluate the DUPE capabilities of LLMs, we create a DUPEd version of the StrategyQA dataset, where facts are manipulated to reverse the answer to the question. Our findings show that even the most advanced GPT models struggle to reason on manipulated facts - showcasing poor DUPE skills - with accuracy dropping by 45% compared to the original dataset. We also investigate prompt settings inspired from student simulation models, which mitigate the accuracy drop to some extent. Our findings have practical implications for understanding the performance of LLMs in real-world applications such as student simulation models that involve reasoning over inaccurate information.

Many recent developments in large language models focus on prompting them to perform specific tasks. One effective prompting method is in-context learning, where the model performs a (possibly new) generation/prediction task given one (or more) examples. Past work has shown that the choice of examples can make a large impact on task performance. However, finding good examples is not straightforward since the definition of a representative group of examples can vary greatly depending on the task. While there are many existing methods for selecting in-context examples, they generally score examples independently, ignoring the dependency between them and the order in which they are provided to the large language model. In this work, we propose Retrieval for In-Context Learning (RetICL), a learnable method for modeling and optimally selecting examples sequentially for in-context learning. We frame the problem of sequential example selection as a Markov decision process, design an example retriever model using an LSTM, and train it using proximal policy optimization (PPO). We validate RetICL on math problem solving datasets and show that it outperforms both heuristic and learnable baselines, and achieves state-of-the-art accuracy on the TabMWP dataset. We also use case studies to show that RetICL implicitly learns representations of math problem solving strategies.

Representing texts as information about entities has long been deemed effective in event reasoning. We propose OpenPI2.0, an improved dataset for tracking entity states in procedural texts. OpenPI2.0 features not only canonicalized entities that facilitate evaluation, but also salience annotations including both manual labels and automatic predictions. Regarding entity salience, we provide a survey on annotation subjectivity, modeling feasibility, and downstream applications in tasks such as question answering and classical planning.

Style is used to convey authors' intentions and attitudes. Despite the success of large pre-trained language models on style classification, prior work relies on fine-tuning with labeled examples. Prompting large language models to classify style without fine-tuning is challenging because language styles can be difficult to define. In this study, we investigate the effectiveness of style lexicons as a means for instructing language models how to identify new styles that are unseen during training. Our experiments show that lexicon-based instructions improve transfer zero-shot performance significantly. We will release our code and data.

The quality of texts generated by natural language generation (NLG) systems is hard to measure automatically. Conventional reference-based metrics, such as BLEU and ROUGE, have been shown to have relatively low correlation with human judgments, especially for tasks that require creativity and diversity. Recent studies suggest using large language models (LLMs) as reference-free metrics for NLG evaluation, which have the benefit of being applicable to new tasks that lack human references. However, these LLM-based evaluators still have lower human correspondence than medium-size neural evaluators. In this work, we present G-Eval, a framework of using large language models with chain-of-thoughts (CoT) and a form-filling paradigm, to assess the quality of NLG outputs. We experiment with two generation tasks, text summarization and dialogue generation. We show that G-Eval with GPT-4 as the backbone model achieves a Spearman correlation of 0.514 with human on summarization task, outperforming all previous methods by a large margin. We also propose preliminary analysis on the behavior of LLM-based evaluators, and highlight the potential issue of LLM-based evaluators having a bias towards the LLM-generated texts. The code is at https://github.com/nlpyang/geval

Visual language such as charts and plots is ubiquitous in the human world. Comprehending plots and charts requires strong reasoning skills. Prior state-of-the-art (SOTA) models require at least tens of thousands of training examples and their reasoning capabilities are still much limited, especially on complex human-written queries. This paper presents the first one-shot solution to visual language reasoning. We decompose the challenge of visual language reasoning into two steps: (1) plot-to-text translation, and (2) reasoning over the translated text. The key in this method is a modality conversion module, named as DePlot, which translates the image of a plot or chart to a linearized table. The output of DePlot can then be directly used to prompt a pretrained large language model (LLM), exploiting the few-shot reasoning capabilities of LLMs. To obtain DePlot, we standardize the plot-to-table task by establishing unified task formats and metrics, and train DePlot end-to-end on this task. DePlot can then be used off-the-shelf together with LLMs in a plug-and-play fashion. Compared with a SOTA model finetuned on more than >28k data points, DePlot+LLM with just one-shot prompting achieves a 24.0% improvement over finetuned SOTA on human-written queries from the task of chart QA.

Large pre-trained language models (PLMs) have made significant progress in encoding world knowledge and spawned a new set of learning paradigms including zero-shot, few-shot, and in-context learning. Many language tasks can be modeled as a set of prompts (for example, is this text about geography?) and language models can provide binary answers, i.e., Yes or No. There is evidence to suggest that the next-word prediction used by many PLMs does not align well with zero-shot paradigms. Therefore, PLMs are fine-tuned as a question-answering system. In-context learning extends zero-shot learning by incorporating prompts and examples, resulting in increased task accuracy. Our paper presents EXnet, a model specifically designed to perform in-context learning without any limitations on the number of examples. We argue that in-context learning is an effective method to increase task accuracy, and providing examples facilitates cross-task generalization, especially when it comes to text classification tasks. With extensive experiments, we show that even our smallest model (15M parameters) generalizes to several unseen classification tasks and domains.

In recent years, the field of computer vision has seen significant advancements thanks to the development of large language models (LLMs). These models have enabled more effective and sophisticated interactions between humans and machines, paving the way for novel techniques that blur the lines between human and machine intelligence. In this paper, we introduce a new paradigm for object detection that we call reasoning-based object detection. Unlike conventional object detection methods that rely on specific object names, our approach enables users to interact with the system using natural language instructions, allowing for a higher level of interactivity. Our proposed method, called DetGPT, leverages state-of-the-art multi-modal models and open-vocabulary object detectors to perform reasoning within the context of the user's instructions and the visual scene. This enables DetGPT to automatically locate the object of interest based on the user's expressed desires, even if the object is not explicitly mentioned. For instance, if a user expresses a desire for a cold beverage, DetGPT can analyze the image, identify a fridge, and use its knowledge of typical fridge contents to locate the beverage. This flexibility makes our system applicable across a wide range of fields, from robotics and automation to autonomous driving. Overall, our proposed paradigm and DetGPT demonstrate the potential for more sophisticated and intuitive interactions between humans and machines. We hope that our proposed paradigm and approach will provide inspiration to the community and open the door to more interative and versatile object detection systems. Our project page is launched at detgpt.github.io.

The recent explosion in work on neural topic modeling has been criticized for optimizing automated topic evaluation metrics at the expense of actual meaningful topic identification. But human annotation remains expensive and time-consuming. We propose LLM-based methods inspired by standard human topic evaluations, in a family of metrics called Contextualized Topic Coherence (CTC). We evaluate both a fully automated version as well as a semi-automated CTC that allows human-centered evaluation of coherence while maintaining the efficiency of automated methods. We evaluate CTC relative to five other metrics on six topic models and find that it outperforms automated topic coherence methods, works well on short documents, and is not susceptible to meaningless but high-scoring topics.

Prompting methods such as Chain-of-Thought (CoT) have shed new light on enhancing the reasoning capabilities of large language models, and researchers have extensively explored the generation process of rationales and answers. However, they have overlooked the potential challenges posed by the poor quality of reasoning problems, which may influence the reasoning performance significantly. In this work, we propose Self-Polish (SP), a novel method that facilitates the model's problem-solving process by prompting them to progressively refine the given problems to be more comprehensible and solvable. Specifically, the method teaches models to eliminate irrelevant information, rearrange the logic structure and organize local conditions into new ones parallelly. SP is orthogonal to all other prompting methods, making it convenient to integrate with state-of-the-art techniques for further improvement. We conduct thorough experiments on five benchmarks to illustrate the effectiveness of the proposed method. For example, with Text-davinci-003, our method boosts the performance of standard few-shot prompting by $8.0\%$ on GSM8K and $17.8\%$ on MultiArith; it also improves the performance of CoT by $6.0\%$ on GSM8K and $6.0\%$ on MathQA, respectively. Furthermore, our method also showcases impressive performance on robustness evaluation.