Authors: Jaroslaw Szumega, Lamine Bougueroua, Blerina Gkotse, Pierre Jouvelot, Federico Ravotti
With the Open Science approach becoming important for research, the evolution towards open scientific-paper reviews is making an impact on the scientific community. However, there is a lack of publicly available resources for conducting research activities related to this subject, as only a limited number of journals and conferences currently allow access to their review process for interested parties. In this paper, we introduce the new comprehensive Open Review-Based dataset (ORB); it includes a curated list of more than 36,000 scientific papers with their more than 89,000 reviews and final decisions. We gather this information from two sources: the OpenReview.net and SciPost.org websites. However, given the volatile nature of this domain, the software infrastructure that we introduce to supplement the ORB dataset is designed to accommodate additional resources in the future. The ORB deliverables include (1) Python code (interfaces and implementations) to translate document data and metadata into a structured and high-level representation, (2) an ETL process (Extract, Transform, Load) to facilitate the automatic updates from defined sources and (3) data files representing the structured data. The paper presents our data architecture and an overview of the collected data along with relevant statistics. For illustration purposes, we also discuss preliminary Natural-Language-Processing-based experiments that aim to predict (1) papers' acceptance based on their textual embeddings, and (2) grading statistics inferred from embeddings as well. We believe ORB provides a valuable resource for researchers interested in open science and review, with our implementation easing the use of this data for further analysis and experimentation. We plan to update ORB as the field matures as well as introduce new resources even more fitted to dedicated scientific domains such as High-Energy Physics.
Authors: Tianyu He, Guanghui Fu, Yijing Yu, Fan Wang, Jianqiang Li, Qing Zhao, Changwei Song, Hongzhi Qi, Dan Luo, Huijing Zou, Bing Xiang Yang
The complexity of psychological principles underscore a significant societal challenge, given the vast social implications of psychological problems. Bridging the gap between understanding these principles and their actual clinical and real-world applications demands rigorous exploration and adept implementation. In recent times, the swift advancement of highly adaptive and reusable artificial intelligence (AI) models has emerged as a promising way to unlock unprecedented capabilities in the realm of psychology. This paper emphasizes the importance of performance validation for these large-scale AI models, emphasizing the need to offer a comprehensive assessment of their verification from diverse perspectives. Moreover, we review the cutting-edge advancements and practical implementations of these expansive models in psychology, highlighting pivotal work spanning areas such as social media analytics, clinical nursing insights, vigilant community monitoring, and the nuanced exploration of psychological theories. Based on our review, we project an acceleration in the progress of psychological fields, driven by these large-scale AI models. These future generalist AI models harbor the potential to substantially curtail labor costs and alleviate social stress. However, this forward momentum will not be without its set of challenges, especially when considering the paradigm changes and upgrades required for medical instrumentation and related applications.
Authors: Montgomery Gole, Williams-Paul Nwadiugwu, Andriy Miranskyy
Sarcasm is a form of irony that requires readers or listeners to interpret its intended meaning by considering context and social cues. Machine learning classification models have long had difficulty detecting sarcasm due to its social complexity and contradictory nature.
This paper explores the applications of the Generative Pretrained Transformer (GPT) models, including GPT-3, InstructGPT, GPT-3.5, and GPT-4, in detecting sarcasm in natural language. It tests fine-tuned and zero-shot models of different sizes and releases.
The GPT models were tested on the political and balanced (pol-bal) portion of the popular Self-Annotated Reddit Corpus (SARC 2.0) sarcasm dataset. In the fine-tuning case, the largest fine-tuned GPT-3 model achieves accuracy and $F_1$-score of 0.81, outperforming prior models. In the zero-shot case, one of GPT-4 models yields an accuracy of 0.70 and $F_1$-score of 0.75. Other models score lower. Additionally, a model's performance may improve or deteriorate with each release, highlighting the need to reassess performance after each release.
Authors: Wannaphong Phatthiyaphaibun, Korakot Chaovavanich, Charin Polpanumas, Arthit Suriyawongkul, Lalita Lowphansirikul, Pattarawat Chormai, Peerat Limkonchotiwat, Thanathip Suntorntip, Can Udomcharoenchaikit
We present PyThaiNLP, a free and open-source natural language processing (NLP) library for Thai language implemented in Python. It provides a wide range of software, models, and datasets for Thai language. We first provide a brief historical context of tools for Thai language prior to the development of PyThaiNLP. We then outline the functionalities it provided as well as datasets and pre-trained language models. We later summarize its development milestones and discuss our experience during its development. We conclude by demonstrating how industrial and research communities utilize PyThaiNLP in their work. The library is freely available at https://github.com/pythainlp/pythainlp.
Authors: Ruoyao Wang, Peter Jansen
In this work, we introduce a self-supervised behavior cloning transformer for text games, which are challenging benchmarks for multi-step reasoning in virtual environments. Traditionally, Behavior Cloning Transformers excel in such tasks but rely on supervised training data. Our approach auto-generates training data by exploring trajectories (defined by common macro-action sequences) that lead to reward within the games, while determining the generality and utility of these trajectories by rapidly training small models then evaluating their performance on unseen development games. Through empirical analysis, we show our method consistently uncovers generalizable training data, achieving about 90\% performance of supervised systems across three benchmark text games.
Authors: Sungryull Sohn, Yiwei Lyu, Anthony Liu, Lajanugen Logeswaran, Dong-Ki Kim, Dongsub Shim, Honglak Lee
Task-Oriented Dialogue (TOD) systems have become crucial components in interactive artificial intelligence applications. While recent advances have capitalized on pre-trained language models (PLMs), they exhibit limitations regarding transparency and controllability. To address these challenges, we propose a novel approach focusing on inferring the TOD-Flow graph from dialogue data annotated with dialog acts, uncovering the underlying task structure in the form of a graph. The inferred TOD-Flow graph can be easily integrated with any dialogue model to improve its prediction performance, transparency, and controllability. Our TOD-Flow graph learns what a model can, should, and should not predict, effectively reducing the search space and providing a rationale for the model's prediction. We show that the proposed TOD-Flow graph better resembles human-annotated graphs compared to prior approaches. Furthermore, when combined with several dialogue policies and end-to-end dialogue models, we demonstrate that our approach significantly improves dialog act classification and end-to-end response generation performance in the MultiWOZ and SGD benchmarks. Code available at: https://github.com/srsohn/TOD-Flow
Authors: Zifan Xu, Haozhu Wang, Dmitriy Bespalov, Peter Stone, Yanjun Qi
Recent advances in Large Language Models (LLMs) have led to an emergent ability of chain-of-thought (CoT) prompting, a prompt reasoning strategy that adds intermediate rationale steps between questions and answers to construct prompts. Conditioned on these prompts, LLMs can effectively learn in context to generate rationales that lead to more accurate answers than when answering the same question directly. To design LLM prompts, one important setting, called demonstration selection, considers selecting demonstrations from an example bank. Existing methods use various heuristics for this selection, but for CoT prompting, which involves unique rationales, it is essential to base the selection upon the intrinsic skills that CoT rationales need, for instance, the skills of addition or subtraction for math word problems.
To address this requirement, we introduce a novel approach named Reasoning Skill Discovery (RSD) that use unsupervised learning to create a latent space representation of rationales, called a reasoning skill. Simultaneously, RSD learns a reasoning policy to determine the required reasoning skill for a given question. This can then guide the selection of examples that demonstrate the required reasoning skills. Our approach offers several desirable properties: it is (1) theoretically grounded, (2) sample-efficient, requiring no LLM inference or manual prompt design, and (3) LLM-agnostic. Empirically, RSD outperforms existing methods by up to 6% in terms of the answer accuracy across multiple reasoning tasks.
Authors: Victor Agostinelli, Max Wild, Matthew Raffel, Kazi Asif Fuad, Lizhong Chen
Large language models (LLMs) with billions of parameters and pretrained on massive amounts of data are now capable of near or better than state-of-the-art performance in a variety of downstream natural language processing tasks. Neural machine translation (NMT) is one such task that LLMs have been applied to with great success. However, little research has focused on applying LLMs to the more difficult subset of NMT called simultaneous translation (SimulMT), where translation begins before the entire source context is available to the model. In this paper, we address key challenges facing LLMs fine-tuned for SimulMT, validate classical SimulMT concepts and practices in the context of LLMs, explore adapting LLMs that are fine-tuned for NMT to the task of SimulMT, and introduce Simul-LLM, the first open-source fine-tuning and evaluation pipeline development framework for LLMs focused on SimulMT.
Authors: Jan Kocoń
Sentiment analysis involves using WordNets enriched with emotional metadata, which are valuable resources. However, manual annotation is time-consuming and expensive, resulting in only a few WordNet Lexical Units being annotated. This paper introduces two new techniques for automatically propagating sentiment annotations from a partially annotated WordNet to its entirety and to a WordNet in a different language: Multilingual Structured Synset Embeddings (MSSE) and Cross-Lingual Deep Neural Sentiment Propagation (CLDNS). We evaluated the proposed MSSE+CLDNS method extensively using Princeton WordNet and Polish WordNet, which have many inter-lingual relations. Our results show that the MSSE+CLDNS method outperforms existing propagation methods, indicating its effectiveness in enriching WordNets with emotional metadata across multiple languages. This work provides a solid foundation for large-scale, multilingual sentiment analysis and is valuable for academic research and practical applications.
Authors: Stanisław Woźniak, Jan Kocoń
In the era of artificial intelligence, data is gold but costly to annotate. The paper demonstrates a groundbreaking solution to this dilemma using ChatGPT for text augmentation in sentiment analysis. We leverage ChatGPT's generative capabilities to create synthetic training data that significantly improves the performance of smaller models, making them competitive with, or even outperforming, their larger counterparts. This innovation enables models to be both efficient and effective, thereby reducing computational cost, inference time, and memory usage without compromising on quality. Our work marks a key advancement in the cost-effective development and deployment of robust sentiment analysis models.
Authors: Sabrina McCallum, Max Taylor-Davies, Stefano V. Albrecht, Alessandro Suglia
Despite numerous successes, the field of reinforcement learning (RL) remains far from matching the impressive generalisation power of human behaviour learning. One possible way to help bridge this gap be to provide RL agents with richer, more human-like feedback expressed in natural language. To investigate this idea, we first extend BabyAI to automatically generate language feedback from the environment dynamics and goal condition success. Then, we modify the Decision Transformer architecture to take advantage of this additional signal. We find that training with language feedback either in place of or in addition to the return-to-go or goal descriptions improves agents' generalisation performance, and that agents can benefit from feedback even when this is only available during training, but not at inference.
Authors: Rui Xue, Xipeng Shen, Ruozhou Yu, Xiaorui Liu
Learning from Text-Attributed Graphs (TAGs) has attracted significant attention due to its wide range of real-world applications. The rapid evolution of large language models (LLMs) has revolutionized the way we process textual data, which indicates a strong potential to replace shallow text embedding generally used in Graph Neural Networks (GNNs). However, we find that existing LLM approaches that exploit text information in graphs suffer from inferior computation and data efficiency. In this work, we introduce a novel and efficient approach for the end-to-end fine-tuning of Large Language Models (LLMs) on TAGs, named LEADING. The proposed approach maintains computation cost and memory overhead comparable to the graph-less fine-tuning of LLMs. Moreover, it transfers the rick knowledge in LLMs to downstream graph learning tasks effectively with limited labeled data in semi-supervised learning. Its superior computation and data efficiency are demonstrated through comprehensive experiments, offering a promising solution for a wide range of LLMs and graph learning tasks on TAGs.
Authors: Mehmet Saygin Seyfioglu, Wisdom O. Ikezogwo, Fatemeh Ghezloo, Ranjay Krishna, Linda Shapiro
The gigapixel scale of whole slide images (WSIs) poses a challenge for histopathology multi-modal chatbots, requiring a global WSI analysis for diagnosis, compounding evidence from different WSI patches. Current visual instruction datasets, generated through large language models, focus on creating question/answer pairs for individual image patches, which may lack diagnostic capacity on their own in histopathology, further complicated by the absence of spatial grounding in histopathology image captions. To bridge this gap, we introduce Quilt-Instruct, a large-scale dataset of 107,131 histopathology-specific instruction question/answer pairs, that is collected by leveraging educational histopathology videos from YouTube, which provides spatial localization of captions by automatically extracting narrators' cursor movements. In addition, we provide contextual reasoning by extracting diagnosis and supporting facts from the entire video content to guide the extrapolative reasoning of GPT-4. Using Quilt-Instruct, we train Quilt-LLaVA, which can reason beyond the given single image patch, enabling diagnostic reasoning and the capability of spatial awareness. To evaluate Quilt-LLaVA, we propose a comprehensive evaluation dataset created from 985 images and 1283 human-generated question-answers. We also thoroughly evaluate Quilt-LLaVA using public histopathology datasets, where Quilt-LLaVA significantly outperforms SOTA by over 10% on relative GPT-4 score and 4% and 9% on open and closed set VQA. Our code, data, and model are publicly available at quilt-llava.github.io.
Authors: Shuli Jiang, Swanand Ravindra Kadhe, Yi Zhou, Ling Cai, Nathalie Baracaldo
Growing applications of large language models (LLMs) trained by a third party raise serious concerns on the security vulnerability of LLMs.It has been demonstrated that malicious actors can covertly exploit these vulnerabilities in LLMs through poisoning attacks aimed at generating undesirable outputs. While poisoning attacks have received significant attention in the image domain (e.g., object detection), and classification tasks, their implications for generative models, particularly in the realm of natural language generation (NLG) tasks, remain poorly understood. To bridge this gap, we perform a comprehensive exploration of various poisoning techniques to assess their effectiveness across a range of generative tasks. Furthermore, we introduce a range of metrics designed to quantify the success and stealthiness of poisoning attacks specifically tailored to NLG tasks. Through extensive experiments on multiple NLG tasks, LLMs and datasets, we show that it is possible to successfully poison an LLM during the fine-tuning stage using as little as 1\% of the total tuning data samples. Our paper presents the first systematic approach to comprehend poisoning attacks targeting NLG tasks considering a wide range of triggers and attack settings. We hope our findings will assist the AI security community in devising appropriate defenses against such threats.
Authors: Atnafu Lambebo Tonja, Melkamu Mersha, Ananya Kalita, Olga Kolesnikova, Jugal Kalita
This paper presents the creation of initial bilingual corpora for thirteen very low-resource languages of India, all from Northeast India. It also presents the results of initial translation efforts in these languages. It creates the first-ever parallel corpora for these languages and provides initial benchmark neural machine translation results for these languages. We intend to extend these corpora to include a large number of low-resource Indian languages and integrate the effort with our prior work with African and American-Indian languages to create corpora covering a large number of languages from across the world.
Authors: Jun Bai, Xiaofeng Zhang, Chen Li, Hanhua Hong, Xi Xu, Chenghua Lin, Wenge Rong
Transferability estimation has been attached to great attention in the computer vision fields. Researchers try to estimate with low computational cost the performance of a model when transferred from a source task to a given target task. Considering the effectiveness of such estimations, the communities of natural language processing also began to study similar problems for the selection of pre-trained language models. However, there is a lack of a comprehensive comparison between these estimation methods yet. Also, the differences between vision and language scenarios make it doubtful whether previous conclusions can be established across fields. In this paper, we first conduct a thorough survey of existing transferability estimation methods being able to find the most suitable model, then we conduct a detailed empirical study for the surveyed methods based on the GLUE benchmark. From qualitative and quantitative analyses, we demonstrate the strengths and weaknesses of existing methods and show that H-Score generally performs well with superiorities in effectiveness and efficiency. We also outline the difficulties of consideration of training details, applicability to text generation, and consistency to certain metrics which shed light on future directions.
Authors: Xinchen Yu, Eduardo Blanco, Lingzi Hong
User-generated replies to hate speech are promising means to combat hatred, but questions about whether they can stop incivility in follow-up conversations linger. We argue that effective replies stop incivility from emerging in follow-up conversations - replies that elicit more incivility are counterproductive. This study introduces the task of predicting the incivility of conversations following replies to hate speech. We first propose a metric to measure conversation incivility based on the number of civil and uncivil comments as well as the unique authors involved in the discourse. Our metric approximates human judgments more accurately than previous metrics. We then use the metric to evaluate the outcomes of replies to hate speech. A linguistic analysis uncovers the differences in the language of replies that elicit follow-up conversations with high and low incivility. Experimental results show that forecasting incivility is challenging. We close with a qualitative analysis shedding light into the most common errors made by the best model.
Authors: Ke Wang, Jun Xie, Yuqi Zhang, Yu Zhao
Improving neural machine translation (NMT) systems with prompting has achieved significant progress in recent years. In this work, we focus on how to integrate multi-knowledge, multiple types of knowledge, into NMT models to enhance the performance with prompting. We propose a unified framework, which can integrate effectively multiple types of knowledge including sentences, terminologies/phrases and translation templates into NMT models. We utilize multiple types of knowledge as prefix-prompts of input for the encoder and decoder of NMT models to guide the translation process. The approach requires no changes to the model architecture and effectively adapts to domain-specific translation without retraining. The experiments on English-Chinese and English-German translation demonstrate that our approach significantly outperform strong baselines, achieving high translation quality and terminology match accuracy.
Authors: Boyi Zeng, Chenghu Zhou, Xinbing Wang, Zhouhan Lin
Protecting the copyright of large language models (LLMs) has become crucial due to their resource-intensive training and accompanying carefully designed licenses. However, identifying the original base model of an LLM is challenging due to potential parameter alterations through fine-tuning or continued pretraining. In this study, we introduce HuRef, a human-readable fingerprint for LLMs that uniquely identifies the base model without exposing model parameters or interfering with training. We first observe that the vector direction of LLM parameters remains stable after the model has converged during pretraining, showing negligible perturbations through subsequent training steps, including continued pretraining, supervised fine-tuning (SFT), and RLHF, which makes it a sufficient condition to identify the base model. The necessity is validated by continuing to train an LLM with an extra term to drive away the model parameters' direction and the model becomes damaged. However, this direction is vulnerable to simple attacks like dimension permutation or matrix rotation, which significantly change it without affecting performance. To address this, leveraging the Transformer structure, we systematically analyze potential attacks and define three invariant terms that identify an LLM's base model. We make these invariant terms human-readable by mapping them to a Gaussian vector using a convolutional encoder and then converting it into a natural image with StyleGAN2. Our method generates a dog image as an identity fingerprint for an LLM, where the dog's appearance strongly indicates the LLM's base model. Experimental results across various LLMs demonstrate the effectiveness of our method, the generated dog image remains invariant to different training steps, including SFT, RLHF, or even continued pretraining with augmented vocabulary in a new language.
Authors: Jae Sung Park, Jack Hessel, Khyathi Raghavi Chandu, Paul Pu Liang, Ximing Lu, Peter West, Youngjae Yu, Qiuyuan Huang, Jianfeng Gao, Ali Farhadi, Yejin Choi
Instruction following vision-language (VL) models offer a flexible interface that supports a broad range of multimodal tasks in a zero-shot fashion. However, interfaces that operate on full images do not directly enable the user to "point to" and access specific regions within images. This capability is important not only to support reference-grounded VL benchmarks, but also, for practical applications that require precise within-image reasoning. We build Localized Visual Commonsense models, which allow users to specify (multiple) regions as input. We train our model by sampling localized commonsense knowledge from a large language model (LLM): specifically, we prompt an LLM to collect commonsense knowledge given a global literal image description and a local literal region description automatically generated by a set of VL models. With a separately trained critic model that selects high-quality examples, we find that training on the localized commonsense corpus can successfully distill existing VL models to support a reference-as-input interface. Empirical results and human evaluations in a zero-shot setup demonstrate that our distillation method results in more precise VL models of reasoning compared to a baseline of passing a generated referring expression to an LLM.
Authors: Virgile Rennard, Guokan Shang, Damien Grari, Julie Hunter, Michalis Vazirgiannis
Recent advances in deep learning, and especially the invention of encoder-decoder architectures, has significantly improved the performance of abstractive summarization systems. The majority of research has focused on written documents, however, neglecting the problem of multi-party dialogue summarization. In this paper, we present a dataset of French political debates for the purpose of enhancing resources for multi-lingual dialogue summarization. Our dataset consists of manually transcribed and annotated political debates, covering a range of topics and perspectives. We highlight the importance of high quality transcription and annotations for training accurate and effective dialogue summarization models, and emphasize the need for multilingual resources to support dialogue summarization in non-English languages. We also provide baseline experiments using state-of-the-art methods, and encourage further research in this area to advance the field of dialogue summarization. Our dataset will be made publicly available for use by the research community.
Authors: Haotian Wang, Xiyuan Du, Weijiang Yu, Qianglong Chen, Kun Zhu, Zheng Chu, Lian Yan, Yi Guan
Multi-agent debate systems are designed to derive accurate and consistent conclusions through adversarial interactions among agents. However, these systems often encounter challenges due to cognitive constraints, manifesting as (1) agents' obstinate adherence to incorrect viewpoints and (2) their propensity to abandon correct viewpoints. These issues are primarily responsible for the ineffectiveness of such debates. Addressing the challenge of cognitive constraints, we introduce a novel framework, the Multi-Agent Debate with Retrieval Augmented (MADRA). MADRA incorporates retrieval of prior knowledge into the debate process, effectively breaking cognitive constraints and enhancing the agents' reasoning capabilities. Furthermore, we have developed a self-selection module within this framework, enabling agents to autonomously select pertinent evidence, thereby minimizing the impact of irrelevant or noisy data. We have comprehensively tested and analyzed MADRA across six diverse datasets. The experimental results demonstrate that our approach significantly enhances performance across various tasks, proving the effectiveness of our proposed method.
Authors: Xiaobin Tian, Zequn Sun, Wei Hu
Entity alignment seeks identical entities in different knowledge graphs, which is a long-standing task in the database research. Recent work leverages deep learning to embed entities in vector space and align them via nearest neighbor search. Although embedding-based entity alignment has gained marked success in recent years, it lacks explanations for alignment decisions. In this paper, we present the first framework that can generate explanations for understanding and repairing embedding-based entity alignment results. Given an entity alignment pair produced by an embedding model, we first compare its neighbor entities and relations to build a matching subgraph as a local explanation. We then construct an alignment dependency graph to understand the pair from an abstract perspective. Finally, we repair the pair by resolving three types of alignment conflicts based on dependency graphs. Experiments on five datasets demonstrate the effectiveness and generalization of our framework in explaining and repairing embedding-based entity alignment results.
Authors: Yujie Qian, Zhening Li, Zhengkai Tu, Connor W. Coley, Regina Barzilay
This paper focuses on using natural language descriptions to enhance predictive models in the chemistry field. Conventionally, chemoinformatics models are trained with extensive structured data manually extracted from the literature. In this paper, we introduce TextReact, a novel method that directly augments predictive chemistry with texts retrieved from the literature. TextReact retrieves text descriptions relevant for a given chemical reaction, and then aligns them with the molecular representation of the reaction. This alignment is enhanced via an auxiliary masked LM objective incorporated in the predictor training. We empirically validate the framework on two chemistry tasks: reaction condition recommendation and one-step retrosynthesis. By leveraging text retrieval, TextReact significantly outperforms state-of-the-art chemoinformatics models trained solely on molecular data.
Authors: Kristina Schaaff, Tim Schlippe, Lorenz Mindner
In this paper we analyze features to classify human- and AI-generated text for English, French, German and Spanish and compare them across languages. We investigate two scenarios: (1) The detection of text generated by AI from scratch, and (2) the detection of text rephrased by AI. For training and testing the classifiers in this multilingual setting, we created a new text corpus covering 10 topics for each language. For the detection of AI-generated text, the combination of all proposed features performs best, indicating that our features are portable to other related languages: The F1-scores are close with 99% for Spanish, 98% for English, 97% for German and 95% for French. For the detection of AI-rephrased text, the systems with all features outperform systems with other features in many cases, but using only document features performs best for German (72%) and Spanish (86%) and only text vector features leads to best results for English (78%).
Authors: Haojie Pan, Zepeng Zhai, Hao Yuan, Yaojia Lv, Ruiji Fu, Ming Liu, Zhongyuan Wang, Bing Qin
Driven by curiosity, humans have continually sought to explore and understand the world around them, leading to the invention of various tools to satiate this inquisitiveness. Despite not having the capacity to process and memorize vast amounts of information in their brains, humans excel in critical thinking, planning, reflection, and harnessing available tools to interact with and interpret the world, enabling them to find answers efficiently. The recent advancements in large language models (LLMs) suggest that machines might also possess the aforementioned human-like capabilities, allowing them to exhibit powerful abilities even with a constrained parameter count. In this paper, we introduce KwaiAgents, a generalized information-seeking agent system based on LLMs. Within KwaiAgents, we propose an agent system that employs LLMs as its cognitive core, which is capable of understanding a user's query, behavior guidelines, and referencing external documents. The agent can also update and retrieve information from its internal memory, plan and execute actions using a time-aware search-browse toolkit, and ultimately provide a comprehensive response. We further investigate the system's performance when powered by LLMs less advanced than GPT-4, and introduce the Meta-Agent Tuning (MAT) framework, designed to ensure even an open-sourced 7B or 13B model performs well among many agent systems. We exploit both benchmark and human evaluations to systematically validate these capabilities. Extensive experiments show the superiority of our agent system compared to other autonomous agents and highlight the enhanced generalized agent-abilities of our fine-tuned LLMs.
Authors: Huan Zhao, Qian Ling, Yi Pan, Tianyang Zhong, Jin-Yu Hu, Junjie Yao, Fengqian Xiao, Zhenxiang Xiao, Yutong Zhang, San-Hua Xu, Shi-Nan Wu, Min Kang, Zihao Wu, Zhengliang Liu, Xi Jiang, Tianming Liu, Yi Shao
In recent years, pre-trained large language models (LLMs) have achieved tremendous success in the field of Natural Language Processing (NLP). Prior studies have primarily focused on general and generic domains, with relatively less research on specialized LLMs in the medical field. The specialization and high accuracy requirements for diagnosis in the medical field, as well as the challenges in collecting large-scale data, have constrained the application and development of LLMs in medical scenarios. In the field of ophthalmology, clinical diagnosis mainly relies on doctors' interpretation of reports and making diagnostic decisions. In order to take advantage of LLMs to provide decision support for doctors, we collected three modalities of ophthalmic report data and fine-tuned the LLaMA2 model, successfully constructing an LLM termed the "Ophtha-LLaMA2" specifically tailored for ophthalmic disease diagnosis. Inference test results show that even with a smaller fine-tuning dataset, Ophtha-LLaMA2 performs significantly better in ophthalmic diagnosis compared to other LLMs. It demonstrates that the Ophtha-LLaMA2 exhibits satisfying accuracy and efficiency in ophthalmic disease diagnosis, making it a valuable tool for ophthalmologists to provide improved diagnostic support for patients. This research provides a useful reference for the application of LLMs in the field of ophthalmology, while showcasing the immense potential and prospects in this domain.
Authors: Simran Arora, Sabri Eyuboglu, Aman Timalsina, Isys Johnson, Michael Poli, James Zou, Atri Rudra, Christopher Ré
Attention-free language models that combine gating and convolutions are growing in popularity due to their efficiency and increasingly competitive performance. To better understand these architectures, we pretrain a suite of 17 attention and "gated-convolution" language models, finding that SoTA gated-convolution architectures still underperform attention by up to 2.1 perplexity points on the Pile. In fine-grained analysis, we find 82% of the gap is explained by each model's ability to recall information that is previously mentioned in-context, e.g. "Hakuna Matata means no worries Hakuna Matata it means no" $\rightarrow$ "??". On this task, termed "associative recall", we find that attention outperforms gated-convolutions by a large margin: a 70M parameter attention model outperforms a 1.4 billion parameter gated-convolution model on associative recall. This is surprising because prior work shows gated convolutions can perfectly solve synthetic tests for AR capability. To close the gap between synthetics and real language, we develop a new formalization of the task called multi-query associative recall (MQAR) that better reflects actual language. We perform an empirical and theoretical study of MQAR that elucidates differences in the parameter-efficiency of attention and gated-convolution recall. Informed by our analysis, we evaluate simple convolution-attention hybrids and show that hybrids with input-dependent sparse attention patterns can close 97.4% of the gap to attention, while maintaining sub-quadratic scaling. Our code is accessible at: https://github.com/HazyResearch/zoology.
Authors: Lucas Weber, Elia Bruni, Dieuwke Hupkes
Just like the previous generation of task-tuned models, large language models (LLMs) that are adapted to tasks via prompt-based methods like in-context-learning (ICL) perform well in some setups but not in others. This lack of consistency in prompt-based learning hints at a lack of robust generalisation. We here introduce the ICL consistency test -- a contribution to the GenBench collaborative benchmark task (CBT) -- which evaluates how consistent a model makes predictions across many different setups while using the same data. The test is based on different established natural language inference tasks. We provide preprocessed data constituting 96 different 'setups' and a metric that estimates model consistency across these setups. The metric is provided on a fine-grained level to understand what properties of a setup render predictions unstable and on an aggregated level to compare overall model consistency. We conduct an empirical analysis of eight state-of-the-art models, and our consistency metric reveals how all tested LLMs lack robust generalisation.
Authors: Sihan Xu, Yidong Huang, Jiayi Pan, Ziqiao Ma, Joyce Chai
Despite recent advances in inversion-based editing, text-guided image manipulation remains challenging for diffusion models. The primary bottlenecks include 1) the time-consuming nature of the inversion process; 2) the struggle to balance consistency with accuracy; 3) the lack of compatibility with efficient consistency sampling methods used in consistency models. To address the above issues, we start by asking ourselves if the inversion process can be eliminated for editing. We show that when the initial sample is known, a special variance schedule reduces the denoising step to the same form as the multi-step consistency sampling. We name this Denoising Diffusion Consistent Model (DDCM), and note that it implies a virtual inversion strategy without explicit inversion in sampling. We further unify the attention control mechanisms in a tuning-free framework for text-guided editing. Combining them, we present inversion-free editing (InfEdit), which allows for consistent and faithful editing for both rigid and non-rigid semantic changes, catering to intricate modifications without compromising on the image's integrity and explicit inversion. Through extensive experiments, InfEdit shows strong performance in various editing tasks and also maintains a seamless workflow (less than 3 seconds on one single A40), demonstrating the potential for real-time applications. Project Page: https://sled-group.github.io/InfEdit/
Authors: Yookyung Kho, Jaehee Kim, Pilsung Kang
Recently, prompt-based fine-tuning has garnered considerable interest as a core technique for few-shot text classification task. This approach reformulates the fine-tuning objective to align with the Masked Language Modeling (MLM) objective. Leveraging unlabeled data, prompt-based self-training has shown greater effectiveness in binary and three-class classification. However, prompt-based self-training for multi-class classification has not been adequately investigated, despite its significant applicability to real-world scenarios. Moreover, extending current methods to multi-class classification suffers from the verbalizer that extracts the predicted value of manually pre-defined single label word for each class from MLM predictions. Consequently, we introduce a novel, efficient verbalizer structure, named Mapping-free Automatic Verbalizer (MAV). Comprising two fully connected layers, MAV serves as a trainable verbalizer that automatically extracts the requisite word features for classification by capitalizing on all available information from MLM predictions. Experimental results on five multi-class classification datasets indicate MAV's superior self-training efficacy.
Authors: Gaurav Kolhatkar, Akshit Madan, Nidhi Kowtal, Satyajit Roy, Sheetal Sonawane
The conversion of user epics or stories into their appropriate representation in pseudocode or code is a time-consuming task, which can take up a large portion of the time in an industrial project. With this research paper, we aim to present a methodology to generate pseudocode from a given agile user story of small functionalities so as to reduce the overall time spent on the industrial project. Pseudocode is a programming language agnostic representation of the steps involved in a computer program, which can be easily converted into any programming language. Leveraging the potential of Natural Language Processing, we want to simplify the development process in organizations that use the Agile Model of Software Development. We present a methodology to convert a problem described in the English language into pseudocode. This methodology divides the Text to Pseudocode conversion task into two stages or subtasks, each of which is treated like an individual machine translation task. Stage 1 is Text to Code Conversion and Stage 2 is Code to Pseudocode Conversion. We find that the CodeT5 model gives the best results in terms of BLEU score when trained separately on the two subtasks mentioned above. BLEU score is a metric that is used to measure the similarity between a machine-translated text and a set of reference translations.
Authors: Lena Held, Ivan Habernal
Why does an argument end up in the final court decision? Was it deliberated or questioned during the oral hearings? Was there something in the hearings that triggered a particular judge to write a dissenting opinion? Despite the availability of the final judgments of the European Court of Human Rights (ECHR), none of these legal research questions can currently be answered as the ECHR's multilingual oral hearings are not transcribed, structured, or speaker-attributed. We address this fundamental gap by presenting LaCour!, the first corpus of textual oral arguments of the ECHR, consisting of 154 full hearings (2.1 million tokens from over 267 hours of video footage) in English, French, and other court languages, each linked to the corresponding final judgment documents. In addition to the transcribed and partially manually corrected text from the video, we provide sentence-level timestamps and manually annotated role and language labels. We also showcase LaCour! in a set of preliminary experiments that explore the interplay between questions and dissenting opinions. Apart from the use cases in legal NLP, we hope that law students or other interested parties will also use LaCour! as a learning resource, as it is freely available in various formats at https://huggingface.co/datasets/TrustHLT/LaCour.
Authors: Tongxin Hu, Zhuang Li, Xin Jin, Lizhen Qu, Xin Zhang
Annually, e-commerce platforms incur substantial financial losses due to trademark infringements, making it crucial to identify and mitigate potential legal risks tied to merchant information registered to the platforms. However, the absence of high-quality datasets hampers research in this area. To address this gap, our study introduces TMID, a novel dataset to detect trademark infringement in merchant registrations. This is a real-world dataset sourced directly from Alipay, one of the world's largest e-commerce and digital payment platforms. As infringement detection is a legal reasoning task requiring an understanding of the contexts and legal rules, we offer a thorough collection of legal rules and merchant and trademark-related contextual information with annotations from legal experts. We ensure the data quality by performing an extensive statistical analysis. Furthermore, we conduct an empirical study on this dataset to highlight its value and the key challenges. Through this study, we aim to contribute valuable resources to advance research into legal compliance related to trademark infringement within the e-commerce sphere. The dataset is available at https://github.com/emnlpTMID/emnlpTMID.github.io .
Authors: Tatiana Passali, Efstathios Chatzikyriakidis, Stelios Andreadis, Thanos G. Stavropoulos, Anastasia Matonaki, Anestis Fachantidis, Grigorios Tsoumakas
Long sentences have been a persistent issue in written communication for many years since they make it challenging for readers to grasp the main points or follow the initial intention of the writer. This survey, conducted using the PRISMA guidelines, systematically reviews two main strategies for addressing the issue of long sentences: a) sentence compression and b) sentence splitting. An increased trend of interest in this area has been observed since 2005, with significant growth after 2017. Current research is dominated by supervised approaches for both sentence compression and splitting. Yet, there is a considerable gap in weakly and self-supervised techniques, suggesting an opportunity for further research, especially in domains with limited data. In this survey, we categorize and group the most representative methods into a comprehensive taxonomy. We also conduct a comparative evaluation analysis of these methods on common sentence compression and splitting datasets. Finally, we discuss the challenges and limitations of current methods, providing valuable insights for future research directions. This survey is meant to serve as a comprehensive resource for addressing the complexities of long sentences. We aim to enable researchers to make further advancements in the field until long sentences are no longer a barrier to effective communication.
Authors: Olga Golovneva, Sean O'Brien, Ramakanth Pasunuru, Tianlu Wang, Luke Zettlemoyer, Maryam Fazel-Zarandi, Asli Celikyilmaz
With recent advancements in large language models, methods like chain-of-thought prompting to elicit reasoning chains have been shown to improve results on reasoning tasks. However, tasks that require multiple steps of reasoning still pose significant challenges to state-of-the-art models. Drawing inspiration from the beam search algorithm, we propose PathFinder, a tree-search-based reasoning path generation approach. It enhances diverse branching and multi-hop reasoning through the integration of dynamic decoding, enabled by varying sampling methods and parameters. Using constrained reasoning, PathFinder integrates novel quality constraints, pruning, and exploration methods to enhance the efficiency and the quality of generation. Moreover, it includes scoring and ranking features to improve candidate selection. Our approach outperforms competitive baselines on three complex arithmetic and commonsense reasoning tasks by 6% on average. Our model generalizes well to longer, unseen reasoning chains, reflecting similar complexities to beam search with large branching factors.
Authors: Seamless Communication, Loïc Barrault, Yu-An Chung, Mariano Coria Meglioli, David Dale, Ning Dong, Mark Duppenthaler, Paul-Ambroise Duquenne, Brian Ellis, Hady Elsahar, Justin Haaheim, John Hoffman, Min-Jae Hwang, Hirofumi Inaguma, Christopher Klaiber, Ilia Kulikov, Pengwei Li, Daniel Licht, Jean Maillard, Ruslan Mavlyutov, Alice Rakotoarison, Kaushik Ram Sadagopan, Abinesh Ramakrishnan, Tuan Tran, Guillaume Wenzek, Yilin Yang, Ethan Ye, Ivan Evtimov, Pierre Fernandez, Cynthia Gao, Prangthip Hansanti, Elahe Kalbassi, Amanda Kallet, Artyom Kozhevnikov, Gabriel Mejia Gonzalez, Robin San Roman, Christophe Touret, Corinne Wong, Carleigh Wood, Bokai Yu, Pierre Andrews, Can Balioglu, Peng-Jen Chen, Marta R. Costa-jussà, Maha Elbayad, Hongyu Gong, Francisco Guzmán, Kevin Heffernan, et al. (17 additional authors not shown)
Large-scale automatic speech translation systems today lack key features that help machine-mediated communication feel seamless when compared to human-to-human dialogue. In this work, we introduce a family of models that enable end-to-end expressive and multilingual translations in a streaming fashion. First, we contribute an improved version of the massively multilingual and multimodal SeamlessM4T model-SeamlessM4T v2. This newer model, incorporating an updated UnitY2 framework, was trained on more low-resource language data. SeamlessM4T v2 provides the foundation on which our next two models are initiated. SeamlessExpressive enables translation that preserves vocal styles and prosody. Compared to previous efforts in expressive speech research, our work addresses certain underexplored aspects of prosody, such as speech rate and pauses, while also preserving the style of one's voice. As for SeamlessStreaming, our model leverages the Efficient Monotonic Multihead Attention mechanism to generate low-latency target translations without waiting for complete source utterances. As the first of its kind, SeamlessStreaming enables simultaneous speech-to-speech/text translation for multiple source and target languages. To ensure that our models can be used safely and responsibly, we implemented the first known red-teaming effort for multimodal machine translation, a system for the detection and mitigation of added toxicity, a systematic evaluation of gender bias, and an inaudible localized watermarking mechanism designed to dampen the impact of deepfakes. Consequently, we bring major components from SeamlessExpressive and SeamlessStreaming together to form Seamless, the first publicly available system that unlocks expressive cross-lingual communication in real-time. The contributions to this work are publicly released and accessible at https://github.com/facebookresearch/seamless_communication
Authors: Mobashir Sadat, Zhengyu Zhou, Lukas Lange, Jun Araki, Arsalan Gundroo, Bingqing Wang, Rakesh R Menon, Md Rizwan Parvez, Zhe Feng
Hallucination is a well-known phenomenon in text generated by large language models (LLMs). The existence of hallucinatory responses is found in almost all application scenarios e.g., summarization, question-answering (QA) etc. For applications requiring high reliability (e.g., customer-facing assistants), the potential existence of hallucination in LLM-generated text is a critical problem. The amount of hallucination can be reduced by leveraging information retrieval to provide relevant background information to the LLM. However, LLMs can still generate hallucinatory content for various reasons (e.g., prioritizing its parametric knowledge over the context, failure to capture the relevant information from the context, etc.). Detecting hallucinations through automated methods is thus paramount. To facilitate research in this direction, we introduce a sophisticated dataset, DelucionQA, that captures hallucinations made by retrieval-augmented LLMs for a domain-specific QA task. Furthermore, we propose a set of hallucination detection methods to serve as baselines for future works from the research community. Analysis and case study are also provided to share valuable insights on hallucination phenomena in the target scenario.
Authors: Navapat Nananukul, Mayank Kejriwal
Recent progress in generative AI, including large language models (LLMs) like ChatGPT, has opened up significant opportunities in fields ranging from natural language processing to knowledge discovery and data mining. However, there is also a growing awareness that the models can be prone to problems such as making information up or `hallucinations', and faulty reasoning on seemingly simple problems. Because of the popularity of models like ChatGPT, both academic scholars and citizen scientists have documented hallucinations of several different types and severity. Despite this body of work, a formal model for describing and representing these hallucinations (with relevant meta-data) at a fine-grained level, is still lacking. In this paper, we address this gap by presenting the Hallucination Ontology or HALO, a formal, extensible ontology written in OWL that currently offers support for six different types of hallucinations known to arise in LLMs, along with support for provenance and experimental metadata. We also collect and publish a dataset containing hallucinations that we inductively gathered across multiple independent Web sources, and show that HALO can be successfully used to model this dataset and answer competency questions.
Authors: Zhiting Hu, Tianmin Shu
Despite their tremendous success in many applications, large language models often fall short of consistent reasoning and planning in various (language, embodied, and social) scenarios, due to inherent limitations in their inference, learning, and modeling capabilities. In this position paper, we present a new perspective of machine reasoning, LAW, that connects the concepts of Language models, Agent models, and World models, for more robust and versatile reasoning capabilities. In particular, we propose that world and agent models are a better abstraction of reasoning, that introduces the crucial elements of deliberate human-like reasoning, including beliefs about the world and other agents, anticipation of consequences, goals/rewards, and strategic planning. Crucially, language models in LAW serve as a backend to implement the system or its elements and hence provide the computational power and adaptability. We review the recent studies that have made relevant progress and discuss future research directions towards operationalizing the LAW framework.
Authors: Hui Wang, Anh Dang, Zihao Wu, Son Mac
The advancements in Artificial Intelligence (AI) technologies such as ChatGPT have gained popularity in recent days. The integration of ChatGPT in educational contexts has already created attractions due to a wide range of applications. However, the automatic generation of human-like texts also poses potential risks to academic integrity, especially when faced with writing-intensive language courses. Considering the ongoing debates, this study aims to investigate the academic policies and guidelines established by US universities regarding the use of ChatGPT in teaching and learning. The data sources include academic policies, statements, guidelines as well as relevant resources that were provided by the top 50 universities in the United States, according to U.S. News. Thematic analysis and qualitative analysis were employed in the analysis and showed that most top 50 universities were open but cautious towards the integration of generative AI in teaching and learning and also expressed their concerns on ethical usage, accuracy, and data privacy. Most universities also provided a variety of resources and guidelines, including syllabus templates/samples, workshops and discussions, shared articles, and one-on-one consultations, with focuses on general technical introduction, ethical concerns, pedagogical applications, preventive strategies, data privacy, limitations, and detective tools. The findings will inform future policy-making regarding the integration of ChatGPT in college-level education and influence the provision of supportive resources by universities for the appropriate application of ChatGPT in education.
Authors: Jared Edward Reser
This article provides an analytical framework for how to simulate human-like thought processes within a computer. It describes how attention and memory should be structured, updated, and utilized to search for associative additions to the stream of thought. The focus is on replicating the mammalian working memory system, which features two forms of persistent activity: sustained firing (preserving information on the order of seconds) and synaptic potentiation (preserving information from minutes to hours). The article uses a series of over 40 original figures to systematically demonstrate how the iterative updating of these working memory stores provides functional structure to thought and consciousness. In an AI implementation, these two stores should be updated continuously and in an iterative fashion, meaning each state should preserve a proportion of the coactive representations from the state before it. Thus, the set of concepts in working memory will evolve gradually and incrementally over time. This makes each state a revised iteration of the preceding state and causes successive states to overlap and blend with respect to the information they contain. Transitions between states happen as persistent activity spreads activation energy throughout the hierarchical network searching long-term memory for the most appropriate representation to be added to the global workspace. The result is a chain of associatively linked intermediate states capable of advancing toward a solution or goal. Iterative updating is conceptualized here as an information processing strategy, a model of working memory, a theory of consciousness, and an algorithm for designing and programming artificial general intelligence.
Authors: Daniele Comi, Dimitrios Christofidellis, Pier Francesco Piazza, Matteo Manica
Intent discovery is a crucial task in natural language processing, and it is increasingly relevant for various of industrial applications. Identifying novel, unseen intents from user inputs remains one of the biggest challenges in this field. Herein, we propose Zero-Shot-BERT-Adapters, a two-stage method for multilingual intent discovery relying on a Transformer architecture, fine-tuned with Adapters. We train the model for Natural Language Inference (NLI) and later perform unknown intent classification in a zero-shot setting for multiple languages. In our evaluation, we first analyze the quality of the model after adaptive fine-tuning on known classes. Secondly, we evaluate its performance in casting intent classification as an NLI task. Lastly, we test the zero-shot performance of the model on unseen classes, showing how Zero-Shot-BERT-Adapters can effectively perform intent discovery by generating semantically similar intents, if not equal, to the ground-truth ones. Our experiments show how Zero-Shot-BERT-Adapters outperforms various baselines in two zero-shot settings: known intent classification and unseen intent discovery. The proposed pipeline holds the potential for broad application in customer care. It enables automated dynamic triage using a lightweight model that can be easily deployed and scaled in various business scenarios, unlike large language models. Zero-Shot-BERT-Adapters represents an innovative multi-language approach for intent discovery, enabling the online generation of novel intents. A Python package implementing the pipeline and the new datasets we compiled are available at the following link: https://github.com/GT4SD/zero-shot-bert-adapters.
Authors: Sanket Vaibhav Mehta, Jai Gupta, Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Jinfeng Rao, Marc Najork, Emma Strubell, Donald Metzler
Differentiable Search Indices (DSIs) encode a corpus of documents in model parameters and use the same model to answer user queries directly. Despite the strong performance of DSI models, deploying them in situations where the corpus changes over time is computationally expensive because reindexing the corpus requires re-training the model. In this work, we introduce DSI++, a continual learning challenge for DSI to incrementally index new documents while being able to answer queries related to both previously and newly indexed documents. Across different model scales and document identifier representations, we show that continual indexing of new documents leads to considerable forgetting of previously indexed documents. We also hypothesize and verify that the model experiences forgetting events during training, leading to unstable learning. To mitigate these issues, we investigate two approaches. The first focuses on modifying the training dynamics. Flatter minima implicitly alleviate forgetting, so we optimize for flatter loss basins and show that the model stably memorizes more documents ($+12\%$). Next, we introduce a generative memory to sample pseudo-queries for documents and supplement them during continual indexing to prevent forgetting for the retrieval task. Extensive experiments on novel continual indexing benchmarks based on Natural Questions (NQ) and MS MARCO demonstrate that our proposed solution mitigates forgetting significantly. Concretely, it improves the average Hits@10 by $+21.1\%$ over competitive baselines for NQ and requires $6$ times fewer model updates compared to re-training the DSI model for incrementally indexing five corpora in a sequence.
Authors: Maximilian M. Rabe, Dario Paape, Daniela Mertzen, Shravan Vasishth, Ralf Engbert
Models of eye-movement control during reading, developed largely within psychology, usually focus on visual, attentional, lexical, and motor processes but neglect post-lexical language processing; by contrast, models of sentence comprehension processes, developed largely within psycholinguistics, generally focus only on post-lexical language processes. We present a model that combines these two research threads, by integrating eye-movement control and sentence processing. Developing such an integrated model is extremely challenging and computationally demanding, but such an integration is an important step toward complete mathematical models of natural language comprehension in reading. We combine the SWIFT model of eye-movement control (Seelig et al., 2020, doi:10.1016/j.jmp.2019.102313) with key components of the Lewis and Vasishth sentence processing model (Lewis & Vasishth, 2005, doi:10.1207/s15516709cog0000_25). This integration becomes possible, for the first time, due in part to recent advances in successful parameter identification in dynamical models, which allows us to investigate profile log-likelihoods for individual model parameters. We present a fully implemented proof-of-concept model demonstrating how such an integrated model can be achieved; our approach includes Bayesian model inference with Markov Chain Monte Carlo (MCMC) sampling as a key computational tool. The integrated Sentence-Processing and Eye-Movement Activation-Coupled Model (SEAM) can successfully reproduce eye movement patterns that arise due to similarity-based interference in reading. To our knowledge, this is the first-ever integration of a complete process model of eye-movement control with linguistic dependency completion processes in sentence comprehension. In future work, this proof of concept model will need to be evaluated using a comprehensive set of benchmark data.
Authors: Qinghua Lu, Liming Zhu, Xiwei Xu, Zhenchang Xing, Jon Whittle
The release of ChatGPT has drawn huge interests on foundations models. There is a broad consensus that foundations models will be the fundamental building blocks for future AI systems. However, there is a lack of systematic guidance on the architecture design. Particularly, the the rapidly growing capabilities of foundations models can eventually absorb other components of AI systems, posing challenges of moving boundary and interface evolution in architecture design. Furthermore, incorporating foundations models into AI systems raises significant concerns about responsible AI due to their opaque nature and rapidly advancing intelligence. To address these challenges, the paper first presents an architecture evolution of AI systems in the era of foundation models, transitioning from "foundation-model-as-a-connector" to "foundation-model-as-a-monolithic architecture". The paper then identifies key design decisions and proposes a pattern-oriented reference architecture for designing responsible foundation-model-based systems. The patterns can enable the potential of foundation models while minimising associated risks.
Authors: Arseny Moskvichev, Ky-Vinh Mai
We propose a new large-scale (nearly a million questions) ultra-long-context (more than 50,000 words average document length) reading comprehension dataset. Using GPT 3.5, we summarized each scene in 1,500 hand-curated fiction books from Project Gutenberg, which resulted in approximately 150 scene-level summaries per book. After that, we created a number of reading comprehension questions based on these summaries, including three types of multiple-choice scene recognition questions, as well as free-form narrative reconstruction questions. With 990,595 total questions, our dataset is an order of magnitude larger than the closest alternatives. Crucially, most questions have a known ``retention demand'', indicating how long-term of a memory is needed to answer them, which should aid long-term memory performance evaluation. We validate our data in four small-scale experiments: one with human labelers, and three with existing language models. We show that our questions 1) adequately represent the source material 2) can be used to diagnose a model's memory capacity 3) are not trivial for modern language models even when the memory demand does not exceed those models' context lengths. Lastly, we provide our code which can be used to further expand the dataset with minimal human labor.
Authors: Lean Wang, Lei Li, Damai Dai, Deli Chen, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers' processing; (2) the consolidated information in label words serves as a reference for LLMs' final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies.
Authors: Anmol Kabra, Ethan R. Elenberg
Large, general purpose language models have demonstrated impressive performance across many different conversational domains. While multi-domain language models achieve low overall perplexity, their outputs are not guaranteed to stay within the domain of a given input prompt. This paper proposes domain privacy as a novel way to quantify how likely a conditional language model will leak across domains. We also develop policy functions based on token-level domain classification, and propose an efficient fine-tuning method to improve the trained model's domain privacy. Experiments on membership inference attacks show that our proposed method has comparable resiliency to methods adapted from recent literature on differentially private language models.
Authors: Kexun Zhang, Danqing Wang, Jingtao Xia, William Yang Wang, Lei Li
Large language models (LLMs) excel at implementing code from functionality descriptions but struggle with algorithmic problems that require not only implementation but also identification of the suitable algorithm. Moreover, LLM-generated programs lack guaranteed correctness and require human verification. To address these challenges, we propose ALGO, a framework that synthesizes Algorithmic programs with LLM-Generated Oracles to guide the generation and verify their correctness. ALGO first generates a reference oracle by prompting an LLM to exhaustively enumerate all the combinations of relevant variables. This oracle is then utilized to guide an arbitrary search strategy in exploring the algorithm space and to verify the synthesized algorithms. Our study shows that the LLM-generated oracles are correct for 88% of the cases. With the oracles as verifiers, ALGO can be integrated with any existing code generation model in a model-agnostic manner to enhance its performance. Experiments show that when equipped with ALGO, we achieve an 8x better one-submission pass rate over the Codex model and a 2.6x better one-submission pass rate over CodeT, the current state-of-the-art model on CodeContests. We can also get 1.3x better pass rate over the ChatGPT Code Interpreter on unseen problems. The problem set we used for testing, the prompts we used, the verifier and solution programs, and the test cases generated by ALGO are available at https://github.com/zkx06111/ALGO.
Authors: Minje Choi, Jiaxin Pei, Sagar Kumar, Chang Shu, David Jurgens
Large language models (LLMs) have been shown to perform well at a variety of syntactic, discourse, and reasoning tasks. While LLMs are increasingly deployed in many forms including conversational agents that interact with humans, we lack a grounded benchmark to measure how well LLMs understand \textit{social} language. Here, we introduce a new theory-driven benchmark, SocKET, that contains 58 NLP tasks testing social knowledge which we group into five categories: humor & sarcasm, offensiveness, sentiment & emotion, and trustworthiness. In tests on the benchmark, we demonstrate that current models attain only moderate performance but reveal significant potential for task transfer among different types and categories of tasks, which were predicted from theory. Through zero-shot evaluations, we show that pretrained models already possess some innate but limited capabilities of social language understanding and training on one category of tasks can improve zero-shot testing on others. Our benchmark provides a systematic way to analyze model performance on an important dimension of language and points to clear room for improvement to build more socially-aware LLMs. The associated resources are released at https://github.com/minjechoi/SOCKET.
Authors: Davoud Ataee Tarzanagh, Yingcong Li, Xuechen Zhang, Samet Oymak
Attention mechanism is a central component of the transformer architecture which led to the phenomenal success of large language models. However, the theoretical principles underlying the attention mechanism are poorly understood, especially its nonconvex optimization dynamics. In this work, we explore the seminal softmax-attention model $f(\boldsymbol{X})=\langle \boldsymbol{Xv}, \texttt{softmax}(\boldsymbol{XWp})\rangle$, where $\boldsymbol{X}$ is the token sequence and $(\boldsymbol{v},\boldsymbol{W},\boldsymbol{p})$ are trainable parameters. We prove that running gradient descent on $\boldsymbol{p}$, or equivalently $\boldsymbol{W}$, converges in direction to a max-margin solution that separates $\textit{locally-optimal}$ tokens from non-optimal ones. This clearly formalizes attention as an optimal token selection mechanism. Remarkably, our results are applicable to general data and precisely characterize $\textit{optimality}$ of tokens in terms of the value embeddings $\boldsymbol{Xv}$ and problem geometry. We also provide a broader regularization path analysis that establishes the margin maximizing nature of attention even for nonlinear prediction heads. When optimizing $\boldsymbol{v}$ and $\boldsymbol{p}$ simultaneously with logistic loss, we identify conditions under which the regularization paths directionally converge to their respective hard-margin SVM solutions where $\boldsymbol{v}$ separates the input features based on their labels. Interestingly, the SVM formulation of $\boldsymbol{p}$ is influenced by the support vector geometry of $\boldsymbol{v}$. Finally, we verify our theoretical findings via numerical experiments and provide insights.
Authors: Mingyuan Fan, Chengyu Wang, Cen Chen, Yang Liu, Jun Huang
Diffusion models and large language models have emerged as leading-edge generative models, revolutionizing various aspects of human life. However, the practical implementations of these models have also exposed inherent risks, bringing to the forefront their evil sides and sparking concerns regarding their trustworthiness. Despite the wealth of literature on this subject, a comprehensive survey specifically delving into the intersection of large-scale generative models and their trustworthiness remains largely absent. To bridge this gap, this paper investigates both the long-standing and emerging threats associated with these models across four fundamental dimensions: 1) privacy, 2) security, 3) fairness, and 4) responsibility. Based on the investigation results, we develop an extensive map outlining the trustworthiness of large generative models. After that, we provide practical recommendations and potential research directions for future secure applications equipped with large generative models, ultimately promoting the trustworthiness of the models and benefiting the society as a whole.
Authors: Sarah Schwettmann, Tamar Rott Shaham, Joanna Materzynska, Neil Chowdhury, Shuang Li, Jacob Andreas, David Bau, Antonio Torralba
Labeling neural network submodules with human-legible descriptions is useful for many downstream tasks: such descriptions can surface failures, guide interventions, and perhaps even explain important model behaviors. To date, most mechanistic descriptions of trained networks have involved small models, narrowly delimited phenomena, and large amounts of human labor. Labeling all human-interpretable sub-computations in models of increasing size and complexity will almost certainly require tools that can generate and validate descriptions automatically. Recently, techniques that use learned models in-the-loop for labeling have begun to gain traction, but methods for evaluating their efficacy are limited and ad-hoc. How should we validate and compare open-ended labeling tools? This paper introduces FIND (Function INterpretation and Description), a benchmark suite for evaluating the building blocks of automated interpretability methods. FIND contains functions that resemble components of trained neural networks, and accompanying descriptions of the kind we seek to generate. The functions span textual and numeric domains, and involve a range of real-world complexities. We evaluate methods that use pretrained language models (LMs) to produce descriptions of function behavior in natural language and code. Additionally, we introduce a new interactive method in which an Automated Interpretability Agent (AIA) generates function descriptions. We find that an AIA, built from an LM with black-box access to functions, can infer function structure, acting as a scientist by forming hypotheses, proposing experiments, and updating descriptions in light of new data. However, AIA descriptions tend to capture global function behavior and miss local details. These results suggest that FIND will be useful for evaluating more sophisticated interpretability methods before they are applied to real-world models.
Authors: Chengyuan Liu, Fubang Zhao, Lizhi Qing, Yangyang Kang, Changlong Sun, Kun Kuang, Fei Wu
Large Language Models (LLMs) presents significant priority in text understanding and generation. However, LLMs suffer from the risk of generating harmful contents especially while being employed to applications. There are several black-box attack methods, such as Prompt Attack, which can change the behaviour of LLMs and induce LLMs to generate unexpected answers with harmful contents. Researchers are interested in Prompt Attack and Defense with LLMs, while there is no publicly available dataset with high successful attacking rate to evaluate the abilities of defending prompt attack. In this paper, we introduce a pipeline to construct high-quality prompt attack samples, along with a Chinese prompt attack dataset called CPAD. Our prompts aim to induce LLMs to generate unexpected outputs with several carefully designed prompt attack templates and widely concerned attacking contents. Different from previous datasets involving safety estimation, we construct the prompts considering three dimensions: contents, attacking methods and goals. Especially, the attacking goals indicate the behaviour expected after successfully attacking the LLMs, thus the responses can be easily evaluated and analysed. We run several popular Chinese LLMs on our dataset, and the results show that our prompts are significantly harmful to LLMs, with around 70% attack success rate to GPT-3.5. CPAD is publicly available at https://github.com/liuchengyuan123/CPAD.
Authors: Bochuan Cao, Yuanpu Cao, Lu Lin, Jinghui Chen
Recently, Large Language Models (LLMs) have made significant advancements and are now widely used across various domains. Unfortunately, there has been a rising concern that LLMs can be misused to generate harmful or malicious content. Though a line of research has focused on aligning LLMs with human values and preventing them from producing inappropriate content, such alignments are usually vulnerable and can be bypassed by alignment-breaking attacks via adversarially optimized or handcrafted jailbreaking prompts. In this work, we introduce a Robustly Aligned LLM (RA-LLM) to defend against potential alignment-breaking attacks. RA-LLM can be directly constructed upon an existing aligned LLM with a robust alignment checking function, without requiring any expensive retraining or fine-tuning process of the original LLM. Furthermore, we also provide a theoretical analysis for RA-LLM to verify its effectiveness in defending against alignment-breaking attacks. Through real-world experiments on open-source large language models, we demonstrate that RA-LLM can successfully defend against both state-of-the-art adversarial prompts and popular handcrafted jailbreaking prompts by reducing their attack success rates from nearly 100% to around 10% or less.
Authors: Catherine Kosten, Philippe Cudré-Mauroux, Kurt Stockinger
With the recent spike in the number and availability of Large Language Models (LLMs), it has become increasingly important to provide large and realistic benchmarks for evaluating Knowledge Graph Question Answering (KGQA) systems. So far the majority of benchmarks rely on pattern-based SPARQL query generation approaches. The subsequent natural language (NL) question generation is conducted through crowdsourcing or other automated methods, such as rule-based paraphrasing or NL question templates. Although some of these datasets are of considerable size, their pitfall lies in their pattern-based generation approaches, which do not always generalize well to the vague and linguistically diverse questions asked by humans in real-world contexts. In this paper, we introduce Spider4SPARQL - a new SPARQL benchmark dataset featuring 9,693 previously existing manually generated NL questions and 4,721 unique, novel, and complex SPARQL queries of varying complexity. In addition to the NL/SPARQL pairs, we also provide their corresponding 166 knowledge graphs and ontologies, which cover 138 different domains. Our complex benchmark enables novel ways of evaluating the strengths and weaknesses of modern KGQA systems. We evaluate the system with state-of-the-art KGQA systems as well as LLMs, which achieve only up to 45\% execution accuracy, demonstrating that Spider4SPARQL is a challenging benchmark for future research.
Authors: Mahyar Abbasian, Iman Azimi, Amir M. Rahmani, Ramesh Jain
Conversational Health Agents (CHAs) are interactive systems that provide healthcare services, such as assistance, self-awareness, and diagnosis. Current CHAs, especially those utilizing Large Language Models (LLMs), primarily focus on conversation aspects. However, they offer limited agent capabilities specifically lacking multi-step problem-solving, empathetic conversations, and multimodal data analysis. Our aim is to overcome these limitations. In this paper, we propose an LLM-powered framework to empower CHAs to generate a personalized response for users' healthcare queries. This framework provides critical thinking, knowledge acquisition, and problem-solving abilities by integrating healthcare data sources, enabling multilingual and multimodal conversations, and interacting with various user data analysis tools. We illustrate the framework's proficiency in handling complex healthcare tasks via a case study on stress level estimation, showcasing the agent's cognitive and operational capabilities. Powered by our framework, the CHA can provide appropriate responses, when the user inquires about their stress level. To achieve this, it learns to collect photoplethysmogram signals, converts them into heart rate variability, and interprets them as indicators of stress levels.
Authors: Yazhou Zhang, Mengyao Wang, Prayag Tiwari, Qiuchi Li, Benyou Wang, Jing Qin
Large language models (LLMs) and their variants have shown extraordinary efficacy across numerous downstream natural language processing (NLP) tasks, which has presented a new vision for the development of NLP. Despite their remarkable performance in natural language generating (NLG), LLMs lack a distinct focus on the emotion understanding domain. As a result, using LLMs for emotion recognition may lead to suboptimal and inadequate precision. Another limitation of LLMs is that they are typical trained without leveraging multi-modal information. To overcome these limitations, we propose DialogueLLM, a context and emotion knowledge tuned LLM that is obtained by fine-tuning LLaMA models with 13,638 multi-modal (i.e., texts and videos) emotional dialogues. The visual information is considered as the supplementary knowledge to construct high-quality instructions. We offer a comprehensive evaluation of our proposed model on three benchmarking emotion recognition in conversations (ERC) datasets and compare the results against the SOTA baselines and other SOTA LLMs. Additionally, DialogueLLM-7B can be easily trained using LoRA on a 40GB A100 GPU in 5 hours, facilitating reproducibility for other researchers.
Authors: Sheng Lu, Shan Chen, Yingya Li, Danielle Bitterman, Guergana Savova, Iryna Gurevych
In-context learning (ICL) is a new learning paradigm that has gained popularity along with the development of large language models. In this work, we adapt a recently proposed hardness metric, pointwise $\mathcal{V}$-usable information (PVI), to an in-context version (in-context PVI). Compared to the original PVI, in-context PVI is more efficient in that it requires only a few exemplars and does not require fine-tuning. We conducted a comprehensive empirical analysis to evaluate the reliability of in-context PVI. Our findings indicate that in-context PVI estimates exhibit similar characteristics to the original PVI. Specific to the in-context setting, we show that in-context PVI estimates remain consistent across different exemplar selections and numbers of shots. The variance of in-context PVI estimates across different exemplar selections is insignificant, which suggests that in-context PVI are stable. Furthermore, we demonstrate how in-context PVI can be employed to identify challenging instances. Our work highlights the potential of in-context PVI and provides new insights into the capabilities of ICL.
Authors: Herbie Bradley, Andrew Dai, Hannah Teufel, Jenny Zhang, Koen Oostermeijer, Marco Bellagente, Jeff Clune, Kenneth Stanley, Grégory Schott, Joel Lehman
In many text-generation problems, users may prefer not only a single response, but a diverse range of high-quality outputs from which to choose. Quality-diversity (QD) search algorithms aim at such outcomes, by continually improving and diversifying a population of candidates. However, the applicability of QD to qualitative domains, like creative writing, has been limited by the difficulty of algorithmically specifying measures of quality and diversity. Interestingly, recent developments in language models (LMs) have enabled guiding search through AI feedback, wherein LMs are prompted in natural language to evaluate qualitative aspects of text. Leveraging this development, we introduce Quality-Diversity through AI Feedback (QDAIF), wherein an evolutionary algorithm applies LMs to both generate variation and evaluate the quality and diversity of candidate text. When assessed on creative writing domains, QDAIF covers more of a specified search space with high-quality samples than do non-QD controls. Further, human evaluation of QDAIF-generated creative texts validates reasonable agreement between AI and human evaluation. Our results thus highlight the potential of AI feedback to guide open-ended search for creative and original solutions, providing a recipe that seemingly generalizes to many domains and modalities. In this way, QDAIF is a step towards AI systems that can independently search, diversify, evaluate, and improve, which are among the core skills underlying human society's capacity for innovation.
Authors: Microsoft Research AI4Science, Microsoft Azure Quantum
In recent years, groundbreaking advancements in natural language processing have culminated in the emergence of powerful large language models (LLMs), which have showcased remarkable capabilities across a vast array of domains, including the understanding, generation, and translation of natural language, and even tasks that extend beyond language processing. In this report, we delve into the performance of LLMs within the context of scientific discovery, focusing on GPT-4, the state-of-the-art language model. Our investigation spans a diverse range of scientific areas encompassing drug discovery, biology, computational chemistry (density functional theory (DFT) and molecular dynamics (MD)), materials design, and partial differential equations (PDE). Evaluating GPT-4 on scientific tasks is crucial for uncovering its potential across various research domains, validating its domain-specific expertise, accelerating scientific progress, optimizing resource allocation, guiding future model development, and fostering interdisciplinary research. Our exploration methodology primarily consists of expert-driven case assessments, which offer qualitative insights into the model's comprehension of intricate scientific concepts and relationships, and occasionally benchmark testing, which quantitatively evaluates the model's capacity to solve well-defined domain-specific problems. Our preliminary exploration indicates that GPT-4 exhibits promising potential for a variety of scientific applications, demonstrating its aptitude for handling complex problem-solving and knowledge integration tasks. Broadly speaking, we evaluate GPT-4's knowledge base, scientific understanding, scientific numerical calculation abilities, and various scientific prediction capabilities.
Authors: Shitao Xiao, Zheng Liu, Peitian Zhang, Xingrun Xing
The pre-trained language models are continually fine-tuned to better support downstream applications. However, this operation may result in significant performance degeneration on general tasks beyond the targeted domain. To overcome this problem, we propose LM-Cocktail which enables the fine-tuned model to stay resilient in general perspectives. Our method is conducted in the form of model merging, where the fine-tuned language model is merged with the pre-trained base model or the peer models from other domains through weighted average. Despite simplicity, LM-Cocktail is surprisingly effective: the resulted model is able to achieve a strong empirical performance in the whole scope of general tasks while preserving a superior capacity in its targeted domain. We conduct comprehensive experiments with LLama and BGE model on popular benchmarks, including FLAN, MMLU, MTEB, whose results validate the efficacy of our proposed method. The code and checkpoints are available at https://github.com/FlagOpen/FlagEmbedding/tree/master/LM_Cocktail.
Authors: Milan Straka
We present CorPipe, the winning entry to the CRAC 2023 Shared Task on Multilingual Coreference Resolution. Our system is an improved version of our earlier multilingual coreference pipeline, and it surpasses other participants by a large margin of 4.5 percent points. CorPipe first performs mention detection, followed by coreference linking via an antecedent-maximization approach on the retrieved spans. Both tasks are trained jointly on all available corpora using a shared pretrained language model. Our main improvements comprise inputs larger than 512 subwords and changing the mention decoding to support ensembling. The source code is available at https://github.com/ufal/crac2023-corpipe.
Authors: Nianwen Si, Hao Zhang, Heyu Chang, Wenlin Zhang, Dan Qu, Weiqiang Zhang
In recent years, large language models (LLMs) have spurred a new research paradigm in natural language processing. Despite their excellent capability in knowledge-based question answering and reasoning, their potential to retain faulty or even harmful knowledge poses risks of malicious application. The challenge of mitigating this issue and transforming these models into purer assistants is crucial for their widespread applicability. Unfortunately, Retraining LLMs repeatedly to eliminate undesirable knowledge is impractical due to their immense parameters. Knowledge unlearning, derived from analogous studies on machine unlearning, presents a promising avenue to address this concern and is notably advantageous in the context of LLMs. It allows for the removal of harmful knowledge in an efficient manner, without affecting unrelated knowledge in the model. To this end, we provide a survey of knowledge unlearning in the era of LLMs. Firstly, we formally define the knowledge unlearning problem and distinguish it from related works. Subsequently, we categorize existing knowledge unlearning methods into three classes: those based on parameter optimization, parameter merging, and in-context learning, and introduce details of these unlearning methods. We further present evaluation datasets used in existing methods, and finally conclude this survey by presenting the ongoing challenges and future directions.
Authors: Abhay Kumar Yadav, Arjun Singh
In this paper, we introduce a novel fine-tuning technique for language models, which involves incorporating symmetric noise into the embedding process. This method aims to enhance the model's function by more stringently regulating its local curvature, demonstrating superior performance over the current method, NEFTune. When fine-tuning the LLaMA-2-7B model using Alpaca, standard techniques yield a 29.79% score on AlpacaEval. However, our approach, SymNoise, increases this score significantly to 69.04%, using symmetric noisy embeddings. This is a 6.7% improvement over the state-of-the-art method, NEFTune~(64.69%). Furthermore, when tested on various models and stronger baseline instruction datasets, such as Evol-Instruct, ShareGPT, OpenPlatypus, SymNoise consistently outperforms NEFTune. The current literature, including NEFTune, has underscored the importance of more in-depth research into the application of noise-based strategies in the fine-tuning of language models. Our approach, SymNoise, is another significant step towards this direction, showing notable improvement over the existing state-of-the-art method.
Authors: Simon Lermen, Ondřej Kvapil
There has been increasing interest in evaluations of language models for a variety of risks and characteristics. Evaluations relying on natural language understanding for grading can often be performed at scale by using other language models. We test the robustness of these model-graded evaluations to injections on different datasets including a new Deception Eval. These injections resemble direct communication between the testee and the evaluator to change their grading. We extrapolate that future, more intelligent models might manipulate or cooperate with their evaluation model. We find significant susceptibility to these injections in state-of-the-art commercial models on all examined evaluations. Furthermore, similar injections can be used on automated interpretability frameworks to produce misleading model-written explanations. The results inspire future work and should caution against unqualified trust in evaluations and automated interpretability.
Authors: Maia Kotelanski, Robert Gallo, Ashwin Nayak, Thomas Savage
Large Language Models have difficulty communicating uncertainty, which is a significant obstacle to applying LLMs to complex medical tasks. This study evaluates methods to measure LLM confidence when suggesting a diagnosis for challenging clinical vignettes. GPT4 was asked a series of challenging case questions using Chain of Thought and Self Consistency prompting. Multiple methods were investigated to assess model confidence and evaluated on their ability to predict the models observed accuracy. The methods evaluated were Intrinsic Confidence, SC Agreement Frequency and CoT Response Length. SC Agreement Frequency correlated with observed accuracy, yielding a higher Area under the Receiver Operating Characteristic Curve compared to Intrinsic Confidence and CoT Length analysis. SC agreement is the most useful proxy for model confidence, especially for medical diagnosis. Model Intrinsic Confidence and CoT Response Length exhibit a weaker ability to differentiate between correct and incorrect answers, preventing them from being reliable and interpretable markers for model confidence. We conclude GPT4 has a limited ability to assess its own diagnostic accuracy. SC Agreement Frequency is the most useful method to measure GPT4 confidence.
Authors: Nuo Chen, Ning Wu, Shining Liang, Ming Gong, Linjun Shou, Dongmei Zhang, Jia Li
This paper presents an in-depth analysis of Large Language Models (LLMs), focusing on LLaMA, a prominent open-source foundational model in natural language processing. Instead of assessing LLaMA through its generative output, we design multiple-choice tasks to probe its intrinsic understanding in high-order tasks such as reasoning and computation. We examine the model horizontally, comparing different sizes, and vertically, assessing different layers. We unveil several key and uncommon findings based on the designed probing tasks: (1) Horizontally, enlarging model sizes almost could not automatically impart additional knowledge or computational prowess. Instead, it can enhance reasoning abilities, especially in math problem solving, and helps reduce hallucinations, but only beyond certain size thresholds; (2) In vertical analysis, the lower layers of LLaMA lack substantial arithmetic and factual knowledge, showcasing logical thinking, multilingual and recognitive abilities, with top layers housing most computational power and real-world knowledge.
Authors: Chengshu Li, Jacky Liang, Andy Zeng, Xinyun Chen, Karol Hausman, Dorsa Sadigh, Sergey Levine, Li Fei-Fei, Fei Xia, Brian Ichter
Code provides a general syntactic structure to build complex programs and perform precise computations when paired with a code interpreter - we hypothesize that language models (LMs) can leverage code-writing to improve Chain of Thought reasoning not only for logic and arithmetic tasks, but also for semantic ones (and in particular, those that are a mix of both). For example, consider prompting an LM to write code that counts the number of times it detects sarcasm in an essay: the LM may struggle to write an implementation for "detect_sarcasm(string)" that can be executed by the interpreter (handling the edge cases would be insurmountable). However, LMs may still produce a valid solution if they not only write code, but also selectively "emulate" the interpreter by generating the expected output of "detect_sarcasm(string)" and other lines of code that cannot be executed. In this work, we propose Chain of Code (CoC), a simple yet surprisingly effective extension that improves LM code-driven reasoning. The key idea is to encourage LMs to format semantic sub-tasks in a program as flexible pseudocode that the interpreter can explicitly catch undefined behaviors and hand off to simulate with an LM (as an "LMulator"). Experiments demonstrate that Chain of Code outperforms Chain of Thought and other baselines across a variety of benchmarks; on BIG-Bench Hard, Chain of Code achieves 84%, a gain of 12% over Chain of Thought. CoC scales well with large and small models alike, and broadens the scope of reasoning questions that LMs can correctly answer by "thinking in code". Project webpage: https://chain-of-code.github.io.