Authors: Yinheng Li, Shaofei Wang, Han Ding, Hang Chen
Recent advances in large language models (LLMs) have opened new possibilities for artificial intelligence applications in finance. In this paper, we provide a practical survey focused on two key aspects of utilizing LLMs for financial tasks: existing solutions and guidance for adoption.
First, we review current approaches employing LLMs in finance, including leveraging pretrained models via zero-shot or few-shot learning, fine-tuning on domain-specific data, and training custom LLMs from scratch. We summarize key models and evaluate their performance improvements on financial natural language processing tasks.
Second, we propose a decision framework to guide financial professionals in selecting the appropriate LLM solution based on their use case constraints around data, compute, and performance needs. The framework provides a pathway from lightweight experimentation to heavy investment in customized LLMs.
Lastly, we discuss limitations and challenges around leveraging LLMs in financial applications. Overall, this survey aims to synthesize the state-of-the-art and provide a roadmap for responsibly applying LLMs to advance financial AI.
Authors: Nicolas Schwenke, Heinrich Söbke, Eckhard Kraft
The release of the large language model based chatbot ChatGPT in November 2022 has brought considerable attention to the subject of artificial intelligence, not only in the public. From the perspective of higher education, ChatGPT challenges various learning and assessment formats as it significantly reduces the effectiveness of their learning and assessment functionalities. In particular, ChatGPT might be applied to formats that require learners to generate text, such as bachelor theses or student research papers. Accordingly, the research question arises to what extent writing of bachelor theses is still a valid learning and assessment format. Correspondingly, in this study, the first author was asked to write his bachelor's thesis exploiting ChatGPT. For tracing the impact of ChatGPT, methodically an autoethnographic approach was used. First, all considerations on the potential use of ChatGPT were documented in logs and secondly, all ChatGPT chats were logged. Both logs and chat histories were analyzed and are presented along to the recommendations for students regarding the use of ChatGPT suggested by Gimpel et al. (2023). In conclusion, ChatGPT is beneficial in thesis writing during various activities, such as brainstorming, structuring and text revision. However, there arise limitations, e.g., in referencing. Thus, ChatGPT requires a continuous validation of the outcomes generated fostering learning. Currently, ChatGPT is to be valued as a beneficial tool in thesis writing. However, writing a conclusive thesis still requires the learner's meaningful engagement. Accordingly, writing a thesis is still a valid learning and assessment format. With further releases of ChatGPT, an increase in capabilities is to be expected and the research question needs to be reevaluated from time to time.
Authors: Masoud Makrehchi, Dell Zhang, Alina Petrova, John Armour
This is the Proceedings of the 3rd International Workshop on Mining and Learning in the Legal Domain (MLLD-23) which took place in conjunction with the 32nd ACM International Conference on Information and Knowledge Management (CIKM-2023) at the University of Birmingham, Birmingham, UK on Sunday 22nd October 2023.
Authors: Mihran Miroyan, Shiny Weng, Rahul Shah, Lisa Yan, Narges Norouzi
In today's rapidly evolving educational landscape, traditional modes of passive information delivery are giving way to transformative pedagogical approaches that prioritize active student engagement. Within the context of large-scale hybrid classrooms, the challenge lies in fostering meaningful and active interaction between students and course content. This study delves into the significance of measuring students' earnestness during interactive lecture participation exercises. By analyzing students' responses to interactive lecture poll questions, establishing a clear rubric for evaluating earnestness, and conducting a comprehensive assessment, we introduce EIT (Earnest Insight Toolkit), a tool designed to assess students' engagement within interactive lecture participation exercises - particularly in the context of large-scale hybrid classrooms. Through the utilization of EIT, our objective is to equip educators with valuable means of identifying at-risk students for enhancing intervention and support strategies, as well as measuring students' levels of engagement with course content.
Authors: Ashlee Kupor, Candice Morgan, Dorottya Demszky
Providing consistent, individualized feedback to teachers on their instruction can improve student learning outcomes. Such feedback can especially benefit novice instructors who teach on online platforms and have limited access to instructional training. To build scalable measures of instruction, we fine-tune RoBERTa and GPT models to identify five instructional talk moves inspired by accountable talk theory: adding on, connecting, eliciting, probing and revoicing students' ideas. We fine-tune these models on a newly annotated dataset of 2500 instructor utterances derived from transcripts of small group instruction in an online computer science course, Code in Place. Although we find that GPT-3 consistently outperforms RoBERTa in terms of precision, its recall varies significantly. We correlate the instructors' use of each talk move with indicators of student engagement and satisfaction, including students' section attendance, section ratings, and assignment completion rates. We find that using talk moves generally correlates positively with student outcomes, and connecting student ideas has the largest positive impact. These results corroborate previous research on the effectiveness of accountable talk moves and provide exciting avenues for using these models to provide instructors with useful, scalable feedback.
Authors: Yining Ye, Xin Cong, Shizuo Tian, Jiannan Cao, Hao Wang, Yujia Qin, Yaxi Lu, Heyang Yu, Huadong Wang, Yankai Lin, Zhiyuan Liu, Maosong Sun
From ancient water wheels to robotic process automation (RPA), automation technology has evolved throughout history to liberate human beings from arduous tasks. Yet, RPA struggles with tasks needing human-like intelligence, especially in elaborate design of workflow construction and dynamic decision-making in workflow execution. As Large Language Models (LLMs) have emerged human-like intelligence, this paper introduces Agentic Process Automation (APA), a groundbreaking automation paradigm using LLM-based agents for advanced automation by offloading the human labor to agents associated with construction and execution. We then instantiate ProAgent, an LLM-based agent designed to craft workflows from human instructions and make intricate decisions by coordinating specialized agents. Empirical experiments are conducted to detail its construction and execution procedure of workflow, showcasing the feasibility of APA, unveiling the possibility of a new paradigm of automation driven by agents. Our code is public at https://github.com/OpenBMB/ProAgent.
Authors: Andrei Nesterov (1), Laura Hollink (1), Jacco van Ossenbruggen (2) ((1) Centrum Wiskunde & Informatica, (2) VU University Amsterdam)
Web resources in linked open data (LOD) are comprehensible to humans through literal textual values attached to them, such as labels, notes, or comments. Word choices in literals may not always be neutral. When outdated and culturally stereotyping terminology is used in literals, they may appear as offensive to users in interfaces and propagate stereotypes to algorithms trained on them. We study how frequently and in which literals contentious terms about people and cultures occur in LOD and whether there are attempts to mark the usage of such terms. For our analysis, we reuse English and Dutch terms from a knowledge graph that provides opinions of experts from the cultural heritage domain about terms' contentiousness. We inspect occurrences of these terms in four widely used datasets: Wikidata, The Getty Art & Architecture Thesaurus, Princeton WordNet, and Open Dutch WordNet. Some terms are ambiguous and contentious only in particular senses. Applying word sense disambiguation, we generate a set of literals relevant to our analysis. We found that outdated, derogatory, stereotyping terms frequently appear in descriptive and labelling literals, such as preferred labels that are usually displayed in interfaces and used for indexing. In some cases, LOD contributors mark contentious terms with words and phrases in literals (implicit markers) or properties linked to resources (explicit markers). However, such marking is rare and non-consistent in all datasets. Our quantitative and qualitative insights could be helpful in developing more systematic approaches to address the propagation of stereotypes via LOD.
Authors: Florian Baud (LIRIS), Alex Aussem (LIRIS)
Multi-document summarization (MDS) is a difficult task in Natural Language Processing, aiming to summarize information from several documents. However, the source documents are often insufficient to obtain a qualitative summary. We propose a retriever-guided model combined with non-parametric memory for summary generation. This model retrieves relevant candidates from a database and then generates the summary considering the candidates with a copy mechanism and the source documents. The retriever is implemented with Approximate Nearest Neighbor Search (ANN) to search large databases. Our method is evaluated on the MultiXScience dataset which includes scientific articles. Finally, we discuss our results and possible directions for future work.
Authors: Rui Fukushima, Jun Tani
ChatGPT, a widely-recognized large language model (LLM), has recently gained substantial attention for its performance scaling, attributed to the billions of web-sourced natural language sentences used for training. Its underlying architecture, Transformer, has found applications across diverse fields, including video, audio signals, and robotic movement. %The crucial question this raises concerns the Transformer's generalization-in-learning (GIL) capacity. However, this raises a crucial question about Transformer's generalization in learning (GIL) capacity. Is ChatGPT's success chiefly due to the vast dataset used for training, or is there more to the story? To investigate this, we compared Transformer's GIL capabilities with those of a traditional Recurrent Neural Network (RNN) in tasks involving attractor dynamics learning. For performance evaluation, the Dynamic Time Warping (DTW) method has been employed. Our simulation results suggest that under conditions of limited data availability, Transformer's GIL abilities are markedly inferior to those of RNN.
Authors: Yufeng Chen
The challenge of improving translation accuracy in GPT-4 is being addressed by harnessing a method known as in-context learning. This paper introduces a strategic approach to utilize in-context learning specifically for machine translation, aiming to significantly boost accuracy. The crux of this method lies in the judicious selection of demonstrations that are most effective for in-context learning. By selecting these examples carefully, GPT-4 can utilize them to achieve remarkably accurate machine translations, eliminating the need for task-specific fine-tuning. This technique is anchored in the semantic similarities between the user's prompt and the chosen dataset. Sentences from this dataset, carefully picked for their relevance and clarity, serve as potent demonstrations for in-context learning. This approach not only enhances translation accuracy but also enriches the understanding of nuanced linguistic structures. It represents a significant step forward in machine learning, leveraging the inherent capabilities of GPT-4 to provide translations that are not only accurate but also contextually rich and linguistically sophisticated. This method demonstrates the potential of in-context learning in overcoming language barriers, opening new avenues for cross-cultural communication and global collaboration.
Authors: Jing Yao, Xiaoyuan Yi, Xiting Wang, Yifan Gong, Xing Xie
The rapid advancement of Large Language Models (LLMs) has attracted much attention to value alignment for their responsible development. However, how to define values in this context remains a largely unexplored question. Existing work mainly follows the Helpful, Honest, Harmless principle and specifies values as risk criteria formulated in the AI community, e.g., fairness and privacy protection, suffering from poor clarity, adaptability and transparency. Inspired by basic values in humanity and social science across cultures, this work proposes a novel basic value alignment paradigm and introduces a value space spanned by basic value dimensions. All LLMs' behaviors can be mapped into the space by identifying the underlying values, possessing the potential to address the three challenges. To foster future research, we apply the representative Schwartz's Theory of Basic Values as an initialized example and construct FULCRA, a dataset consisting of 5k (LLM output, value vector) pairs. Our extensive analysis of FULCRA reveals the underlying relation between basic values and LLMs' behaviors, demonstrating that our approach not only covers existing mainstream risks but also anticipates possibly unidentified ones. Additionally, we present an initial implementation of the basic value evaluation and alignment, paving the way for future research in this line.
Authors: Cicero Nogueira dos Santos, James Lee-Thorp, Isaac Noble, Chung-Ching Chang, David Uthus
Scaling up the number of parameters of language models has proven to be an effective approach to improve performance. For dense models, increasing model size proportionally increases the model's computation footprint. In this work, we seek to aggressively decouple learning capacity and FLOPs through Mixture-of-Experts (MoE) style models with large knowledge-rich vocabulary based routing functions and experts. Our proposed approach, dubbed Mixture of Word Experts (MoWE), can be seen as a memory augmented model, where a large set of word-specific experts play the role of a sparse memory. We demonstrate that MoWE performs significantly better than the T5 family of models with similar number of FLOPs in a variety of NLP tasks. Additionally, MoWE outperforms regular MoE models on knowledge intensive tasks and has similar performance to more complex memory augmented approaches that often require to invoke custom mechanisms to search the sparse memory.
Authors: Peter Belcak, Roger Wattenhofer
Language models only really need to use an exponential fraction of their neurons for individual inferences. As proof, we present FastBERT, a BERT variant that uses 0.3\% of its neurons during inference while performing on par with similar BERT models. FastBERT selectively engages just 12 out of 4095 neurons for each layer inference. This is achieved by replacing feedforward networks with fast feedforward networks (FFFs). While no truly efficient implementation currently exists to unlock the full acceleration potential of conditional neural execution, we provide high-level CPU code achieving 78x speedup over the optimized baseline feedforward implementation, and a PyTorch implementation delivering 40x speedup over the equivalent batched feedforward inference. We publish our training code, benchmarking setup, and model weights.
Authors: Sara Shatnawi, Sawsan Alqahtani, Hanan Aldarmaki
Automatic text-based diacritic restoration models generally have high diacritic error rates when applied to speech transcripts as a result of domain and style shifts in spoken language. In this work, we explore the possibility of improving the performance of automatic diacritic restoration when applied to speech data by utilizing the parallel spoken utterances. In particular, we use the pre-trained Whisper ASR model fine-tuned on relatively small amounts of diacritized Arabic speech data to produce rough diacritized transcripts for the speech utterances, which we then use as an additional input for a transformer-based diacritic restoration model. The proposed model consistently improve diacritic restoration performance compared to an equivalent text-only model, with at least 5\% absolute reduction in diacritic error rate within the same domain and on two out-of-domain test sets. Our results underscore the inadequacy of current text-based diacritic restoration models for speech data sets and provide a new baseline for speech-based diacritic restoration.
Authors: Fuxiao Liu, Xiaoyang Wang, Wenlin Yao, Jianshu Chen, Kaiqiang Song, Sangwoo Cho, Yaser Yacoob, Dong Yu
With the rapid development of large language models (LLMs) and their integration into large multimodal models (LMMs), there has been impressive progress in zero-shot completion of user-oriented vision-language tasks. However, a gap remains in the domain of chart image understanding due to the distinct abstract components in charts. To address this, we introduce a large-scale MultiModal Chart Instruction (MMC-Instruction) dataset comprising 600k instances supporting diverse tasks and chart types. Leveraging this data, we develop MultiModal Chart Assistant (MMCA), an LMM that achieves state-of-the-art performance on existing chart QA benchmarks. Recognizing the need for a comprehensive evaluation of LMM chart understanding, we also propose a MultiModal Chart Benchmark (MMC-Benchmark), a comprehensive human-annotated benchmark with 9 distinct tasks evaluating reasoning capabilities over charts. Extensive experiments on MMC-Benchmark reveal the limitations of existing LMMs on correctly interpreting charts, even for the most recent GPT-4V model. Our work provides an instruction-tuning methodology and benchmark to advance multimodal understanding of charts.
Authors: Nicholas Farn, Richard Shin
Large language models (LLMs) have displayed massive improvements in reasoning and decision-making skills and can hold natural conversations with users. Many recent works seek to augment LLM-based assistants with external tools so they can access private or up-to-date information and carry out actions on behalf of users. To better measure the performance of these assistants, this paper introduces ToolTalk, a benchmark consisting of complex user intents requiring multi-step tool usage specified through dialogue. ToolTalk contains 28 tools grouped into 7 plugins, and includes a complete simulated implementation of each tool, allowing for fully automated evaluation of assistants that rely on execution feedback. ToolTalk also emphasizes tools that externally affect the world rather than only tools for referencing or searching information. We evaluate GPT-3.5 and GPT-4 on ToolTalk resulting in success rates of 26% and 50% respectively. Our analysis of the errors reveals three major categories and suggests some future directions for improvement. We release ToolTalk at https://github.com/microsoft/ToolTalk.
Authors: Yan Cathy Hua, Paul Denny, Katerina Taskova, Jöerg Wicker
Aspect-based Sentiment Analysis (ABSA) is a type of fine-grained sentiment analysis (SA) that identifies aspects and the associated opinions from a given text. In the digital era, ABSA gained increasing popularity and applications in mining opinionated text data to obtain insights and support decisions. ABSA research employs linguistic, statistical, and machine-learning approaches and utilises resources such as labelled datasets, aspect and sentiment lexicons and ontology. By its nature, ABSA is domain-dependent and can be sensitive to the impact of misalignment between the resource and application domains. However, to our knowledge, this topic has not been explored by the existing ABSA literature reviews. In this paper, we present a Systematic Literature Review (SLR) of ABSA studies with a focus on the research application domain, dataset domain, and the research methods to examine their relationships and identify trends over time. Our results suggest a number of potential systemic issues in the ABSA research literature, including the predominance of the ``product/service review'' dataset domain among the majority of studies that did not have a specific research application domain, coupled with the prevalence of dataset-reliant methods such as supervised machine learning. This review makes a number of unique contributions to the ABSA research field: 1) To our knowledge, it is the first SLR that links the research domain, dataset domain, and research method through a systematic perspective; 2) it is one of the largest scoped SLR on ABSA, with 519 eligible studies filtered from 4191 search results without time constraint; and 3) our review methodology adopted an innovative automatic filtering process based on PDF-mining, which enhanced screening quality and reliability. Suggestions and our review limitations are also discussed.
Authors: Hyundong Cho, Shuai Liu, Taiwei Shi, Darpan Jain, Basem Rizk, Yuyang Huang, Zixun Lu, Nuan Wen, Jonathan Gratch, Emilio Ferrara, Jonathan May
Human moderation of online conversation is essential to maintaining civility and focus in a dialogue, but is challenging to scale and harmful to moderators. The inclusion of sophisticated natural language generation modules as a force multiplier aid moderators is a tantalizing prospect, but adequate evaluation approaches have so far been elusive. In this paper, we establish a systematic definition of conversational moderation effectiveness through a multidisciplinary lens that incorporates insights from social science. We then propose a comprehensive evaluation framework that uses this definition to asses models' moderation capabilities independently of human intervention. With our framework, we conduct the first known study of conversational dialogue models as moderators, finding that appropriately prompted models can provide specific and fair feedback on toxic behavior but struggle to influence users to increase their levels of respect and cooperation.
Authors: Sayan Putatunda, Anwesha Bhowmik, Girish Thiruvenkadam, Rahul Ghosh
According to the literature, Product reviews are an important source of information for customers to support their buying decision. Product reviews improve customer trust and loyalty. Reviews help customers in understanding what other customers think about a particular product and helps in driving purchase decisions. Therefore, for an e-commerce platform it is important to understand the sentiments in customer reviews to understand their products and services, and it also allows them to potentially create positive consumer interaction as well as long lasting relationships. Reviews also provide innovative ways to market the products for an ecommerce company. One such approach is Nudge Marketing. Nudge marketing is a subtle way for an ecommerce company to help their customers make better decisions without hesitation.
Authors: Sabrine Amri, Henri-Cedric Mputu Boleilanga, Esma Aïmeur
ExFake is an explainable fake news detection system based on content and context-level information. It is concerned with the veracity analysis of online posts based on their content, social context (i.e., online users' credibility and historical behaviour), and data coming from trusted entities such as fact-checking websites and named entities. Unlike state-of-the-art systems, an Explainable AI (XAI) assistant is also adopted to help online social networks (OSN) users develop good reflexes when faced with any doubted information that spreads on social networks. The trustworthiness of OSN users is also addressed by assigning a credibility score to OSN users, as OSN users are one of the main culprits for spreading fake news. Experimental analysis on a real-world dataset demonstrates that ExFake significantly outperforms other baseline methods for fake news detection.
Authors: Federico Albanese, Daniel Ciolek, Nicolas D'Ippolito
In the context of information systems, text sanitization techniques are used to identify and remove sensitive data to comply with security and regulatory requirements. Even though many methods for privacy preservation have been proposed, most of them are focused on the detection of entities from specific domains (e.g., credit card numbers, social security numbers), lacking generality and requiring customization for each desirable domain. Moreover, removing words is, in general, a drastic measure, as it can degrade text coherence and contextual information. Less severe measures include substituting a word for a safe alternative, yet it can be challenging to automatically find meaningful substitutions. We present a zero-shot text sanitization technique that detects and substitutes potentially sensitive information using Large Language Models. Our evaluation shows that our method excels at protecting privacy while maintaining text coherence and contextual information, preserving data utility for downstream tasks.
Authors: Bibek Upadhayay, Vahid Behzadan
LLMs such as ChatGPT and PaLM can be utilized to train on a new language and revitalize low-resource languages. However, it is evidently very costly to pretrain pr fine-tune LLMs to adopt new languages. Another challenge is the limitation of benchmark datasets and the metrics used to measure the performance of models in multilingual settings. This paper proposes cost-effective solutions to both of the aforementioned challenges. We introduce the Multilingual Instruction-Tuning Dataset (MITS), which is comprised of the translation of Alpaca-52K, Dolly-15K, and Vicuna Benchmark in 132 languages. Also, we propose a new method called \emph{TaCo: Translation-Assisted Cross-Linguality}, which make uses of translation in a chain-of-thought process to instruction-tune LLMs on a new languages through a curriculum learning process. As a proof of concept, we experimented with the instruction-tuned Guanaco-33B model and performed further instruction tuning using the TaCo method in three low-resource languages and one high-resource language. Our results show that the TaCo method impresses the GPT-4 with 82% for a low-resource language in the Vicuna Benchmark dataset, and boosts performance by double in contrast to the performance of instruction tuning only. Our results show that TaCo is a promising method for creating multilingual LLMs, even for low-resource languages. We have released our datasets and the model adapters, and encourage the research community to make use of these resources towards advancing work on multilingual LLMs.
Authors: Mathias Vogel
This report explores the challenge of enhancing expressiveness control in Text-to-Speech (TTS) models by augmenting a frozen pretrained model with a Diffusion Model that is conditioned on joint semantic audio/text embeddings. The paper identifies the challenges encountered when working with a VAE-based TTS model and evaluates different image-to-image methods for altering latent speech features. Our results offer valuable insights into the complexities of adding expressiveness control to TTS systems and open avenues for future research in this direction.
Authors: Yao-Shun Chuang, Xiaoqian Jiang, Chun-Teh Lee, Ryan Brandon, Duong Tran, Oluwabunmi Tokede, Muhammad F. Walji
This study explored the usability of prompt generation on named entity recognition (NER) tasks and the performance in different settings of the prompt. The prompt generation by GPT-J models was utilized to directly test the gold standard as well as to generate the seed and further fed to the RoBERTa model with the spaCy package. In the direct test, a lower ratio of negative examples with higher numbers of examples in prompt achieved the best results with a F1 score of 0.72. The performance revealed consistency, 0.92-0.97 in the F1 score, in all settings after training with the RoBERTa model. The study highlighted the importance of seed quality rather than quantity in feeding NER models. This research reports on an efficient and accurate way to mine clinical notes for periodontal diagnoses, allowing researchers to easily and quickly build a NER model with the prompt generation approach.
Authors: Jiageng Mao, Junjie Ye, Yuxi Qian, Marco Pavone, Yue Wang
Human-level driving is an ultimate goal of autonomous driving. Conventional approaches formulate autonomous driving as a perception-prediction-planning framework, yet their systems do not capitalize on the inherent reasoning ability and experiential knowledge of humans. In this paper, we propose a fundamental paradigm shift from current pipelines, exploiting Large Language Models (LLMs) as a cognitive agent to integrate human-like intelligence into autonomous driving systems. Our approach, termed Agent-Driver, transforms the traditional autonomous driving pipeline by introducing a versatile tool library accessible via function calls, a cognitive memory of common sense and experiential knowledge for decision-making, and a reasoning engine capable of chain-of-thought reasoning, task planning, motion planning, and self-reflection. Powered by LLMs, our Agent-Driver is endowed with intuitive common sense and robust reasoning capabilities, thus enabling a more nuanced, human-like approach to autonomous driving. We evaluate our approach on the large-scale nuScenes benchmark, and extensive experiments substantiate that our Agent-Driver significantly outperforms the state-of-the-art driving methods by a large margin. Our approach also demonstrates superior interpretability and few-shot learning ability to these methods. Project page: \href{https://github.com/USC-GVL/Agent-Driver/blob/main/index.html}{here}.
Authors: Joshua Belofsky
This paper introduces a method for adapting LoRA adapters in smaller-sized language models to arbitrary downstream tasks. Unlike standard mixture-of-expert architectures, our method employs a gradient-free routing function to choose a weighted combination of experts without increasing the compute requirements for training or inference. The results show that token-level adaptation of LoRA adapters outperforms the base Llama-2-7b model across mathematical (GSM8K), scientific (ARC-Challenge), reading comprehension (SQuAD), and coding (CodeAlpaca-20k) tasks. Further evaluations also show that the average performance of token-level adaptation outperforms individual models fine-tuned for each of the tasks with the best performance observed in adaptation of every-other token during inference. The code for this study is made available through a public repository.
Authors: Sergei O. Kuznetsov, Vasilii A. Gromov, Nikita S. Borodin, Andrei M. Divavin
Some results of a computational experiment for determining the intrinsic dimension of linguistic varieties for the Bengali and Russian languages are presented. At the same time, both sets of words and sets of bigrams in these languages were considered separately. The method used to solve this problem was based on formal concept analysis algorithms. It was found that the intrinsic dimensions of these languages are significantly less than the dimensions used in popular neural network models in natural language processing.
Authors: Yimeng Li, Navid Rajabi, Sulabh Shrestha, Md Alimoor Reza, Jana Kosecka
The image annotation stage is a critical and often the most time-consuming part required for training and evaluating object detection and semantic segmentation models. Deployment of the existing models in novel environments often requires detecting novel semantic classes not present in the training data. Furthermore, indoor scenes contain significant viewpoint variations, which need to be handled properly by trained perception models. We propose to leverage the recent advancements in state-of-the-art models for bottom-up segmentation (SAM), object detection (Detic), and semantic segmentation (MaskFormer), all trained on large-scale datasets. We aim to develop a cost-effective labeling approach to obtain pseudo-labels for semantic segmentation and object instance detection in indoor environments, with the ultimate goal of facilitating the training of lightweight models for various downstream tasks. We also propose a multi-view labeling fusion stage, which considers the setting where multiple views of the scenes are available and can be used to identify and rectify single-view inconsistencies. We demonstrate the effectiveness of the proposed approach on the Active Vision dataset and the ADE20K dataset. We evaluate the quality of our labeling process by comparing it with human annotations. Also, we demonstrate the effectiveness of the obtained labels in downstream tasks such as object goal navigation and part discovery. In the context of object goal navigation, we depict enhanced performance using this fusion approach compared to a zero-shot baseline that utilizes large monolithic vision-language pre-trained models.
Authors: Shaunak Joshi (1), Raghav Gaggar (1) ((1) University of Southern California)
With the increase in video-sharing platforms across the internet, it is difficult for humans to moderate the data for explicit content. Hence, an automated pipeline to scan through video data for explicit content has become the need of the hour. We propose a novel pipeline that uses multi-modal deep learning to first extract the explicit segments of input videos and then summarize their content using text to determine its age appropriateness and age rating. We also evaluate our pipeline's effectiveness in the end using standard metrics.
Authors: Karel D'Oosterlinck, Thomas Demeester, Chris Develder, Christopher Potts
Model interpretability and model editing are crucial goals in the age of large language models. Interestingly, there exists a link between these two goals: if a method is able to systematically edit model behavior with regard to a human concept of interest, this editor method can help make internal representations more interpretable by pointing towards relevant representations and systematically manipulating them.
Authors: Michael A. Hedderich, Jonas Fischer, Dietrich Klakow, Jilles Vreeken
State-of-the-art NLP methods achieve human-like performance on many tasks, but make errors nevertheless. Characterizing these errors in easily interpretable terms gives insight into whether a classifier is prone to making systematic errors, but also gives a way to act and improve the classifier. We propose to discover those patterns of tokens that distinguish correct and erroneous predictions as to obtain global and interpretable descriptions for arbitrary NLP classifiers. We formulate the problem of finding a succinct and non-redundant set of such patterns in terms of the Minimum Description Length principle. Through an extensive set of experiments, we show that our method, Premise, performs well in practice. Unlike existing solutions, it recovers ground truth, even on highly imbalanced data over large vocabularies. In VQA and NER case studies, we confirm that it gives clear and actionable insight into the systematic errors made by NLP classifiers.
Authors: Eunji Lee, Sihyeon Kim, Sundong Kim, Soyeon Jung, Heeja Kim, Meeyoung Cha
The task of assigning internationally accepted commodity codes (aka HS codes) to traded goods is a critical function of customs offices. Like court decisions made by judges, this task follows the doctrine of precedent and can be nontrivial even for experienced officers. Together with the Korea Customs Service (KCS), we propose a first-ever explainable decision supporting model that suggests the most likely subheadings (i.e., the first six digits) of the HS code. The model also provides reasoning for its suggestion in the form of a document that is interpretable by customs officers. We evaluated the model using 5,000 cases that recently received a classification request. The results showed that the top-3 suggestions made by our model had an accuracy of 93.9\% when classifying 925 challenging subheadings. A user study with 32 customs experts further confirmed that our algorithmic suggestions accompanied by explainable reasonings, can substantially reduce the time and effort taken by customs officers for classification reviews.
Authors: Jon Z. Cai, Shafiuddin Rehan Ahmed, Julia Bonn, Kristin Wright-Bettner, Martha Palmer, James H. Martin
In this paper, we introduce CAMRA (Copilot for AMR Annotatations), a cutting-edge web-based tool designed for constructing Abstract Meaning Representation (AMR) from natural language text. CAMRA offers a novel approach to deep lexical semantics annotation such as AMR, treating AMR annotation akin to coding in programming languages. Leveraging the familiarity of programming paradigms, CAMRA encompasses all essential features of existing AMR editors, including example lookup, while going a step further by integrating Propbank roleset lookup as an autocomplete feature within the tool. Notably, CAMRA incorporates AMR parser models as coding co-pilots, greatly enhancing the efficiency and accuracy of AMR annotators. To demonstrate the tool's capabilities, we provide a live demo accessible at: https://camra.colorado.edu
Authors: Shobhit Agarwal, Yevgeniy R. Semenov, William Lotter
Explainability is a longstanding challenge in deep learning, especially in high-stakes domains like healthcare. Common explainability methods highlight image regions that drive an AI model's decision. Humans, however, heavily rely on language to convey explanations of not only "where" but "what". Additionally, most explainability approaches focus on explaining individual AI predictions, rather than describing the features used by an AI model in general. The latter would be especially useful for model and dataset auditing, and potentially even knowledge generation as AI is increasingly being used in novel tasks. Here, we present an explainability strategy that uses a vision-language model to identify language-based descriptors of a visual classification task. By leveraging a pre-trained joint embedding space between images and text, our approach estimates a new classification task as a linear combination of words, resulting in a weight for each word that indicates its alignment with the vision-based classifier. We assess our approach using two medical imaging classification tasks, where we find that the resulting descriptors largely align with clinical knowledge despite a lack of domain-specific language training. However, our approach also identifies the potential for 'shortcut connections' in the public datasets used. Towards a functional measure of explainability, we perform a pilot reader study where we find that the AI-identified words can enable non-expert humans to perform a specialized medical task at a non-trivial level. Altogether, our results emphasize the potential of using multimodal foundational models to deliver intuitive, language-based explanations of visual tasks.
Authors: Jing Yang Lee, Kong Aik Lee, Woon-Seng Gan
Despite recent progress in generative open-domain dialogue, the issue of low response diversity persists. Prior works have addressed this issue via either novel objective functions, alternative learning approaches such as variational frameworks, or architectural extensions such as the Randomized Link (RL) Transformer. However, these approaches typically entail either additional difficulties during training/inference, or a significant increase in model size and complexity. Hence, we propose the \underline{Pa}rtially \underline{Ra}ndomized trans\underline{Former} (PaRaFormer), a simple extension of the transformer which involves freezing the weights of selected layers after random initialization. Experimental results reveal that the performance of the PaRaformer is comparable to that of the aforementioned approaches, despite not entailing any additional training difficulty or increase in model complexity.
Authors: Panfeng Li, Mohamed Abouelenien, Rada Mihalcea
Deception detection is gaining increasing interest due to ethical and security concerns. This paper explores the application of convolutional neural networks for the purpose of multimodal deception detection. We use a dataset built by interviewing 104 subjects about two topics, with one truthful and one falsified response from each subject about each topic. In particular, we make three main contributions. First, we extract linguistic and physiological features from this data to train and construct the neural network models. Second, we propose a fused convolutional neural network model using both modalities in order to achieve an improved overall performance. Third, we compare our new approach with earlier methods designed for multimodal deception detection. We find that our system outperforms regular classification methods; our results indicate the feasibility of using neural networks for deception detection even in the presence of limited amounts of data.
Authors: Jing Yang Lee, Kong Aik Lee, Woon-Seng Gan
To engage human users in meaningful conversation, open-domain dialogue agents are required to generate diverse and contextually coherent dialogue. Despite recent advancements, which can be attributed to the usage of pretrained language models, the generation of diverse and coherent dialogue remains an open research problem. A popular approach to address this issue involves the adaptation of variational frameworks. However, while these approaches successfully improve diversity, they tend to compromise on contextual coherence. Hence, we propose the Bayesian Open-domain Dialogue with Empirical Bayes (BODEB) framework, an empirical bayes framework for constructing an Bayesian open-domain dialogue agent by leveraging pretrained parameters to inform the prior and posterior parameter distributions. Empirical results show that BODEB achieves better results in terms of both diversity and coherence compared to variational frameworks.
Authors: Sohini Roychowdhury
Generative AI has significantly reduced the entry barrier to the domain of AI owing to the ease of use and core capabilities of automation, translation, and intelligent actions in our day to day lives. Currently, Large language models (LLMs) that power such chatbots are being utilized primarily for their automation capabilities for software monitoring, report generation etc. and for specific personalized question answering capabilities, on a limited scope and scale. One major limitation of the currently evolving family of LLMs is 'hallucinations', wherein inaccurate responses are reported as factual. Hallucinations are primarily caused by biased training data, ambiguous prompts and inaccurate LLM parameters, and they majorly occur while combining mathematical facts with language-based context. Thus, monitoring and controlling for hallucinations becomes necessary when designing solutions that are meant for decision makers. In this work we present the three major stages in the journey of designing hallucination-minimized LLM-based solutions that are specialized for the decision makers of the financial domain, namely: prototyping, scaling and LLM evolution using human feedback. These three stages and the novel data to answer generation modules presented in this work are necessary to ensure that the Generative AI chatbots, autonomous reports and alerts are reliable and high-quality to aid key decision-making processes.
Authors: Varun Khurana, Yaman K Singla, Jayakumar Subramanian, Rajiv Ratn Shah, Changyou Chen, Zhiqiang Xu, Balaji Krishnamurthy
The last few years have witnessed great success on image generation, which has crossed the acceptance thresholds of aesthetics, making it directly applicable to personal and commercial applications. However, images, especially in marketing and advertising applications, are often created as a means to an end as opposed to just aesthetic concerns. The goal can be increasing sales, getting more clicks, likes, or image sales (in the case of stock businesses). Therefore, the generated images need to perform well on these key performance indicators (KPIs), in addition to being aesthetically good. In this paper, we make the first endeavor to answer the question of "How can one infuse the knowledge of the end-goal within the image generation process itself to create not just better-looking images but also "better-performing'' images?''. We propose BoigLLM, an LLM that understands both image content and user behavior. BoigLLM knows how an image should look to get a certain required KPI. We show that BoigLLM outperforms 13x larger models such as GPT-3.5 and GPT-4 in this task, demonstrating that while these state-of-the-art models can understand images, they lack information on how these images perform in the real world. To generate actual pixels of behavior-conditioned images, we train a diffusion-based model (BoigSD) to align with a proposed BoigLLM-defined reward. We show the performance of the overall pipeline on two datasets covering two different behaviors: a stock dataset with the number of forward actions as the KPI and a dataset containing tweets with the total likes as the KPI, denoted as BoigBench. To advance research in the direction of utility-driven image generation and understanding, we release BoigBench, a benchmark dataset containing 168 million enterprise tweets with their media, brand account names, time of post, and total likes.
Authors: Duong Tien Pham, Luan Thanh Nguyen
Every human has their own name, a fundamental aspect of their identity and cultural heritage. The name often conveys a wealth of information, including details about an individual's background, ethnicity, and, especially, their gender. By detecting gender through the analysis of names, researchers can unlock valuable insights into linguistic patterns and cultural norms, which can be applied to practical applications. Hence, this work presents a novel dataset for Japanese name gender detection comprising 64,139 full names in romaji, hiragana, and kanji forms, along with their biological genders. Moreover, we propose Gendec, a framework for gender detection from Japanese names that leverages diverse approaches, including traditional machine learning techniques or cutting-edge transfer learning models, to predict the gender associated with Japanese names accurately. Through a thorough investigation, the proposed framework is expected to be effective and serve potential applications in various domains.
Authors: Dongyuan Li, Yusong Wang, Kotaro Funakoshi, Manabu Okumura
Multimodal emotion recognition aims to recognize emotions for each utterance of multiple modalities, which has received increasing attention for its application in human-machine interaction. Current graph-based methods fail to simultaneously depict global contextual features and local diverse uni-modal features in a dialogue. Furthermore, with the number of graph layers increasing, they easily fall into over-smoothing. In this paper, we propose a method for joint modality fusion and graph contrastive learning for multimodal emotion recognition (Joyful), where multimodality fusion, contrastive learning, and emotion recognition are jointly optimized. Specifically, we first design a new multimodal fusion mechanism that can provide deep interaction and fusion between the global contextual and uni-modal specific features. Then, we introduce a graph contrastive learning framework with inter-view and intra-view contrastive losses to learn more distinguishable representations for samples with different sentiments. Extensive experiments on three benchmark datasets indicate that Joyful achieved state-of-the-art (SOTA) performance compared to all baselines.
Authors: Haoran Zhao, Jake Ryland Williams
While Large Language Models (LLMs) become ever more dominant, classic pre-trained word embeddings sustain their relevance through computational efficiency and nuanced linguistic interpretation. Drawing from recent studies demonstrating that the convergence of GloVe and word2vec optimizations all tend towards log-co-occurrence matrix variants, we construct a novel word representation system called Bit-cipher that eliminates the need of backpropagation while leveraging contextual information and hyper-efficient dimensionality reduction techniques based on unigram frequency, providing strong interpretability, alongside efficiency. We use the bit-cipher algorithm to train word vectors via a two-step process that critically relies on a hyperparameter -- bits -- that controls the vector dimension. While the first step trains the bit-cipher, the second utilizes it under two different aggregation modes -- summation or concatenation -- to produce contextually rich representations from word co-occurrences. We extend our investigation into bit-cipher's efficacy, performing probing experiments on part-of-speech (POS) tagging and named entity recognition (NER) to assess its competitiveness with classic embeddings like word2vec and GloVe. Additionally, we explore its applicability in LM training and fine-tuning. By replacing embedding layers with cipher embeddings, our experiments illustrate the notable efficiency of cipher in accelerating the training process and attaining better optima compared to conventional training paradigms. Experiments on the integration of bit-cipher embedding layers with Roberta, T5, and OPT, prior to or as a substitute for fine-tuning, showcase a promising enhancement to transfer learning, allowing rapid model convergence while preserving competitive performance.
Authors: Uwe D. Reichel
The purposes of the CoPaSul toolkit are (1) automatic prosodic annotation and (2) prosodic feature extraction from syllable to utterance level. CoPaSul stands for contour-based, parametric, superpositional intonation stylization. In this framework intonation is represented as a superposition of global and local contours that are described parametrically in terms of polynomial coefficients. On the global level (usually associated but not necessarily restricted to intonation phrases) the stylization serves to represent register in terms of time-varying F0 level and range. On the local level (e.g. accent groups), local contour shapes are described. From this parameterization several features related to prosodic boundaries and prominence can be derived. Furthermore, by coefficient clustering prosodic contour classes can be obtained in a bottom-up way. Next to the stylization-based feature extraction also standard F0 and energy measures (e.g. mean and variance) as well as rhythmic aspects can be calculated. At the current state automatic annotation comprises: segmentation into interpausal chunks, syllable nucleus extraction, and unsupervised localization of prosodic phrase boundaries and prominent syllables. F0 and partly also energy feature sets can be derived for: standard measurements (as median and IQR), register in terms of F0 level and range, prosodic boundaries, local contour shapes, bottom-up derived contour classes, Gestalt of accent groups in terms of their deviation from higher level prosodic units, as well as for rhythmic aspects quantifying the relation between F0 and energy contours and prosodic event rates.
Authors: Xiangyu Peng, Siyan Li, Sarah Wiegreffe, Mark Riedl
Transformer-based language model approaches to automated story generation currently provide state-of-the-art results. However, they still suffer from plot incoherence when generating narratives over time, and critically lack basic commonsense reasoning. Furthermore, existing methods generally focus only on single-character stories, or fail to track characters at all. To improve the coherence of generated narratives and to expand the scope of character-centric narrative generation, we introduce Commonsense-inference Augmented neural StoryTelling (CAST), a framework for introducing commonsense reasoning into the generation process with the option to model the interaction between multiple characters. We find that our CAST method produces significantly more coherent, on-topic, enjoyable and fluent stories than existing models in both the single-character and two-character settings in three storytelling domains.
Authors: Yanhua Xu, Dominik Wojtczak
The rapid mutation of the influenza virus threatens public health. Reassortment among viruses with different hosts can lead to a fatal pandemic. However, it is difficult to detect the original host of the virus during or after an outbreak as influenza viruses can circulate between different species. Therefore, early and rapid detection of the viral host would help reduce the further spread of the virus. We use various machine learning models with features derived from the position-specific scoring matrix (PSSM) and features learned from word embedding and word encoding to infer the origin host of viruses. The results show that the performance of the PSSM-based model reaches the MCC around 95%, and the F1 around 96%. The MCC obtained using the model with word embedding is around 96%, and the F1 is around 97%.
Authors: Yinghao Aaron Li, Cong Han, Nima Mesgarani
Text-to-Speech (TTS) has recently seen great progress in synthesizing high-quality speech owing to the rapid development of parallel TTS systems, but producing speech with naturalistic prosodic variations, speaking styles and emotional tones remains challenging. Moreover, since duration and speech are generated separately, parallel TTS models still have problems finding the best monotonic alignments that are crucial for naturalistic speech synthesis. Here, we propose StyleTTS, a style-based generative model for parallel TTS that can synthesize diverse speech with natural prosody from a reference speech utterance. With novel Transferable Monotonic Aligner (TMA) and duration-invariant data augmentation schemes, our method significantly outperforms state-of-the-art models on both single and multi-speaker datasets in subjective tests of speech naturalness and speaker similarity. Through self-supervised learning of the speaking styles, our model can synthesize speech with the same prosodic and emotional tone as any given reference speech without the need for explicitly labeling these categories.
Authors: Somnath Basu Roy Chowdhury, Nicholas Monath, Avinava Dubey, Amr Ahmed, Snigdha Chaturvedi
Opinion summarization is the task of creating summaries capturing popular opinions from user reviews. In this paper, we introduce Geodesic Summarizer (GeoSumm), a novel system to perform unsupervised extractive opinion summarization. GeoSumm involves an encoder-decoder based representation learning model, that generates representations of text as a distribution over latent semantic units. GeoSumm generates these representations by performing dictionary learning over pre-trained text representations at multiple decoder layers. We then use these representations to quantify the relevance of review sentences using a novel approximate geodesic distance based scoring mechanism. We use the relevance scores to identify popular opinions in order to compose general and aspect-specific summaries. Our proposed model, GeoSumm, achieves state-of-the-art performance on three opinion summarization datasets. We perform additional experiments to analyze the functioning of our model and showcase the generalization ability of {\X} across different domains.
Authors: Alan Ramponi
Italy is characterized by a one-of-a-kind linguistic diversity landscape in Europe, which implicitly encodes local knowledge, cultural traditions, artistic expressions and history of its speakers. However, most local languages and dialects in Italy are at risk of disappearing within few generations. The NLP community has recently begun to engage with endangered languages, including those of Italy. Yet, most efforts assume that these varieties are under-resourced language monoliths with an established written form and homogeneous functions and needs, and thus highly interchangeable with each other and with high-resource, standardized languages. In this paper, we introduce the linguistic context of Italy and challenge the default machine-centric assumptions of NLP for Italy's language varieties. We advocate for a shift in the paradigm from machine-centric to speaker-centric NLP, and provide recommendations and opportunities for work that prioritizes languages and their speakers over technological advances. To facilitate the process, we finally propose building a local community towards responsible, participatory efforts aimed at supporting vitality of languages and dialects of Italy.
Authors: Abdullatif Köksal, Timo Schick, Hinrich Schütze
Few-shot classification has made great strides due to foundation models that, through priming and prompting, are highly effective few-shot learners. However, this approach has high variance both across different sets of few shots (data selection) and across different finetuning runs (run variability). This is problematic not only because it impedes the fair comparison of different approaches, but especially because it makes few-shot learning too unreliable for many real-world applications. To alleviate these issues, we make two contributions for more stable and effective few-shot learning: First, we propose novel ensembling methods and show that they substantially reduce run variability. Second, we introduce a new active learning (AL) criterion for data selection and present the first AL-based approach specifically tailored towards prompt-based learning. In our experiments, we show that our combined method, MEAL (Multiprompt finetuning and prediction Ensembling with Active Learning), improves overall performance of prompt-based finetuning by 2.3 points on five diverse tasks. We publicly share our code and data splits in https://github.com/akoksal/MEAL.
Authors: Megan Leszczynski, Shu Zhang, Ravi Ganti, Krisztian Balog, Filip Radlinski, Fernando Pereira, Arun Tejasvi Chaganty
Recommender systems are ubiquitous yet often difficult for users to control, and adjust if recommendation quality is poor. This has motivated conversational recommender systems (CRSs), with control provided through natural language feedback. However, as with most application domains, building robust CRSs requires training data that reflects system usage$\unicode{x2014}$here conversations with user utterances paired with items that cover a wide range of preferences. This has proved challenging to collect scalably using conventional methods. We address the question of whether it can be generated synthetically, building on recent advances in natural language. We evaluate in the setting of item set recommendation, noting the increasing attention to this task motivated by use cases like music, news, and recipe recommendation. We present TalkTheWalk, which synthesizes realistic high-quality conversational data by leveraging domain expertise encoded in widely available curated item collections, generating a sequence of hypothetical yet plausible item sets, then using a language model to produce corresponding user utterances. We generate over one million diverse playlist curation conversations in the music domain, and show these contain consistent utterances with relevant item sets nearly matching the quality of an existing but small human-collected dataset for this task. We demonstrate the utility of the generated synthetic dataset on a conversational item retrieval task and show that it improves over both unsupervised baselines and systems trained on a real dataset.
Authors: Sang Michael Xie, Shibani Santurkar, Tengyu Ma, Percy Liang
Selecting a suitable pretraining dataset is crucial for both general-domain (e.g., GPT-3) and domain-specific (e.g., Codex) language models (LMs). We formalize this problem as selecting a subset of a large raw unlabeled dataset to match a desired target distribution given unlabeled target samples. Due to the scale and dimensionality of the raw text data, existing methods use simple heuristics or require human experts to manually curate data. Instead, we extend the classic importance resampling approach used in low-dimensions for LM data selection. We propose Data Selection with Importance Resampling (DSIR), an efficient and scalable framework that estimates importance weights in a reduced feature space for tractability and selects data with importance resampling according to these weights. We instantiate the DSIR framework with hashed n-gram features for efficiency, enabling the selection of 100M documents from the full Pile dataset in 4.5 hours. To measure whether hashed n-gram features preserve the aspects of the data that are relevant to the target, we define KL reduction, a data metric that measures the proximity between the selected pretraining data and the target on some feature space. Across 8 data selection methods (including expert selection), KL reduction on hashed n-gram features highly correlates with average downstream accuracy (r=0.82). When selecting data for continued pretraining on a specific domain, DSIR performs comparably to expert curation across 8 target distributions. When pretraining general-domain models (target is Wikipedia and books), DSIR improves over random selection and heuristic filtering baselines by 2-2.5% on the GLUE benchmark. Code is available at https://github.com/p-lambda/dsir.
Authors: Nicolas Gontier, Pau Rodriguez, Issam Laradji, David Vazquez, Christopher Pal
Text-based game environments are challenging because agents must deal with long sequences of text, execute compositional actions using text and learn from sparse rewards. We address these challenges by proposing Language Decision Transformers (LDTs), a framework that is based on transformer language models and decision transformers (DTs). Our LDTs extend DTs with 3 components: (1) exponential tilt to guide the agent towards high obtainable goals, (2) novel goal conditioning methods yielding better results than the traditional return-to-go (sum of all future rewards), and (3) a model of future observations that improves agent performance. LDTs are the first to address offline RL with DTs on these challenging games. Our experiments show that LDTs achieve the highest scores among many different types of agents on some of the most challenging Jericho games, such as Enchanter.
Authors: Chengwei Qin, Aston Zhang, Zhuosheng Zhang, Jiaao Chen, Michihiro Yasunaga, Diyi Yang
Spurred by advancements in scale, large language models (LLMs) have demonstrated the ability to perform a variety of natural language processing (NLP) tasks zero-shot -- i.e., without adaptation on downstream data. Recently, the debut of ChatGPT has drawn a great deal of attention from the natural language processing (NLP) community due to the fact that it can generate high-quality responses to human input and self-correct previous mistakes based on subsequent conversations. However, it is not yet known whether ChatGPT can serve as a generalist model that can perform many NLP tasks zero-shot. In this work, we empirically analyze the zero-shot learning ability of ChatGPT by evaluating it on 20 popular NLP datasets covering 7 representative task categories. With extensive empirical studies, we demonstrate both the effectiveness and limitations of the current version of ChatGPT. We find that ChatGPT performs well on many tasks favoring reasoning capabilities (e.g., arithmetic reasoning) while it still faces challenges when solving specific tasks such as sequence tagging. We additionally provide in-depth analysis through qualitative case studies.
Authors: Chengwei Qin, Qian Li, Ruochen Zhao, Shafiq Joty
Prompt tuning (PT) which only tunes the embeddings of an additional sequence of tokens per task, keeping the pre-trained language model (PLM) frozen, has shown remarkable performance in few-shot learning. Despite this, PT has been shown to rely heavily on good initialization of the prompt embeddings. In this work, we study meta prompt tuning (MPT) to systematically explore how meta-learning can help improve (if it can) cross-task generalization in PT through learning to initialize the prompt embeddings from other relevant tasks. We empirically analyze a representative set of meta learning algorithms in a wide range of adaptation settings with different source/target task configurations on a large set of few-shot tasks. With extensive experiments and analysis, we demonstrate the effectiveness of MPT. We find the improvement to be significant particularly on classification tasks. For other kinds of tasks such as question answering, we observe that while MPT can outperform PT in most cases, it does not always outperform multi-task learning. We further provide an in-depth analysis from the perspective of task similarity.
Authors: Ron Yosef, Yonatan Bitton, Dafna Shahaf
Figures of speech such as metaphors, similes, and idioms are integral parts of human communication. They are ubiquitous in many forms of discourse, allowing people to convey complex, abstract ideas and evoke emotion. As figurative forms are often conveyed through multiple modalities (e.g., both text and images), understanding multimodal figurative language is an important AI challenge, weaving together profound vision, language, commonsense and cultural knowledge.
In this work, we develop the Image Recognition of Figurative Language (IRFL) dataset. We leverage human annotation and an automatic pipeline we created to generate a multimodal dataset, and introduce two novel tasks as a benchmark for multimodal figurative language understanding. We experimented with state-of-the-art vision and language models and found that the best (22%) performed substantially worse than humans (97%). We release our dataset, benchmark, and code, in hopes of driving the development of models that can better understand figurative language.
Authors: Yifan Jiang, Filip Ilievski, Kaixin Ma
Stories about everyday situations are an essential part of human communication, motivating the need to develop AI agents that can reliably understand these stories. Despite the long list of supervised methods for story completion and procedural understanding, current AI has no mechanisms to automatically track and explain procedures in unseen stories. To bridge this gap, we study the ability of AI models to transfer procedural knowledge to novel narrative tasks in a transparent manner. We design LEAP: a comprehensive framework that integrates state-of-the-art modeling architectures, training regimes, and augmentation strategies based on both natural and synthetic stories. To address the lack of densely annotated training data, we devise a robust automatic labeler based on few-shot prompting to enhance the augmented data. Our experiments with in- and out-of-domain tasks reveal insights into the interplay of different architectures, training regimes, and augmentation strategies. LEAP's labeler has a clear positive impact on out-of-domain datasets, while the resulting dense annotation provides native explainability.
Authors: Abdullatif Köksal, Omer Faruk Yalcin, Ahmet Akbiyik, M. Tahir Kilavuz, Anna Korhonen, Hinrich Schütze
Pretrained language models (PLMs) are key components in NLP, but they contain strong social biases. Quantifying these biases is challenging because current methods focusing on fill-the-mask objectives are sensitive to slight changes in input. To address this, we propose a bias probing technique called LABDet, for evaluating social bias in PLMs with a robust and language-agnostic method. For nationality as a case study, we show that LABDet `surfaces' nationality bias by training a classifier on top of a frozen PLM on non-nationality sentiment detection. We find consistent patterns of nationality bias across monolingual PLMs in six languages that align with historical and political context. We also show for English BERT that bias surfaced by LABDet correlates well with bias in the pretraining data; thus, our work is one of the few studies that directly links pretraining data to PLM behavior. Finally, we verify LABDet's reliability and applicability to different templates and languages through an extensive set of robustness checks. We publicly share our code and dataset in https://github.com/akoksal/LABDet.
Authors: James A. Michaelov, Benjamin K. Bergen
Does inverse scaling only occur as a function of model size, or can it also occur over the course of training? We carry out an exploratory study investigating whether the performance of language models on specific tasks can decrease (while general performance remains high) during training on the language modeling task. We find 8 tasks on which Pythia 12B (Biderman et al., 2023) shows decreased performance over the course of training. Five of these tasks (TruthfulQA-MC1, TruthfulQA-MC2, Hindsight Neglect, Memo Trap, and Pattern Match Suppression) additionally show a consistent relationship whereby larger language models show a greater decrease in performance the more they are trained, despite showing standard (positive) scaling overall. This highlights the importance of testing performance at all relevant benchmarks any time models are trained on additional data, even if their overall performance improves
Authors: Ananya Harsh Jha, Tom Sherborne, Evan Pete Walsh, Dirk Groeneveld, Emma Strubell, Iz Beltagy
With the increase in the size of large language models (LLMs), we need compression methods that can reduce the model size while preserving the generality and zero-shot promptability of the model. This goal is more ambitious than the typical compression setup, which reduces the model's size at the expense of specializing it to a specific end-task. To study this, we develop a task-agnostic compression pipeline with a large-scale evaluation comprising language modeling perplexity and 12 zero-shot end-tasks. Our results show that a simple layer-wise pruning followed by continued language model pretraining matches or outperforms three existing state-of-the-art baselines while being 1.5x more computationally efficient. However, unlike typical task-specialized compression, our best-compressed model significantly underperforms a similar-sized model trained from scratch. We posit the half-sized pretrained model as an upper bound for task-agnostic compression and call for future work to bridge this gap under a reasonable token budget. Our findings highlight the inadequacy of existing compression methods for LLMs and establish a requirement for new methods that preserve a model's generality and zero-shot promptability under compression. We release our code and evaluation setup to facilitate reproducibility and help iterate on method design.
Authors: Amirkeivan Mohtashami, Martin Jaggi
While Transformers have shown remarkable success in natural language processing, their attention mechanism's large memory requirements have limited their ability to handle longer contexts. Prior approaches, such as recurrent memory or retrieval-based augmentation, have either compromised the random-access flexibility of attention (i.e., the capability to select any token in the entire context) or relied on separate mechanisms for relevant context retrieval, which may not be compatible with the model's attention. In this paper, we present a novel approach that allows access to the complete context while retaining random-access flexibility, closely resembling running attention on the entire context. Our method uses a landmark token to represent each block of the input and trains the attention to use it for selecting relevant blocks, enabling retrieval of blocks directly through the attention mechanism instead of by relying on a separate mechanism. Our approach seamlessly integrates with specialized data structures and the system's memory hierarchy, enabling processing of arbitrarily long context lengths. We demonstrate that our method can obtain comparable performance with Transformer-XL while significantly reducing the number of retrieved tokens in each step. Finally, we show that fine-tuning LLaMA 7B with our method successfully extends its context length capacity to over 32k tokens, allowing for inference at the context lengths of GPT-4. We release the implementation of landmark attention and the code to reproduce our experiments at https://github.com/epfml/landmark-attention/.
Authors: Shuo Chen, Jindong Gu, Zhen Han, Yunpu Ma, Philip Torr, Volker Tresp
Various adaptation methods, such as LoRA, prompts, and adapters, have been proposed to enhance the performance of pre-trained vision-language models in specific domains. The robustness of these adaptation methods against distribution shifts have not been studied. In this study, we assess the robustness of 11 widely-used adaptation methods across 4 vision-language datasets under multimodal corruptions. Concretely, we introduce 7 benchmark datasets, including 96 visual and 87 textual corruptions, to investigate the robustness of different adaptation methods, the impact of available adaptation examples, and the influence of trainable parameter size during adaptation. Our analysis reveals that: 1) Adaptation methods are more sensitive to text corruptions than visual corruptions. 2) Full fine-tuning does not consistently provide the highest robustness; instead, adapters can achieve better robustness with comparable clean performance. 3) Contrary to expectations, our findings indicate that increasing the number of adaptation data and parameters does not guarantee enhanced robustness; instead it results in even lower robustness. We hope this study could benefit future research in the development of robust multimodal adaptation methods. The benchmark, code, and dataset used in this study can be accessed at https://adarobustness.github.io .
Authors: Yinghao Aaron Li, Cong Han, Vinay S. Raghavan, Gavin Mischler, Nima Mesgarani
In this paper, we present StyleTTS 2, a text-to-speech (TTS) model that leverages style diffusion and adversarial training with large speech language models (SLMs) to achieve human-level TTS synthesis. StyleTTS 2 differs from its predecessor by modeling styles as a latent random variable through diffusion models to generate the most suitable style for the text without requiring reference speech, achieving efficient latent diffusion while benefiting from the diverse speech synthesis offered by diffusion models. Furthermore, we employ large pre-trained SLMs, such as WavLM, as discriminators with our novel differentiable duration modeling for end-to-end training, resulting in improved speech naturalness. StyleTTS 2 surpasses human recordings on the single-speaker LJSpeech dataset and matches it on the multispeaker VCTK dataset as judged by native English speakers. Moreover, when trained on the LibriTTS dataset, our model outperforms previous publicly available models for zero-shot speaker adaptation. This work achieves the first human-level TTS on both single and multispeaker datasets, showcasing the potential of style diffusion and adversarial training with large SLMs. The audio demos and source code are available at https://styletts2.github.io/.
Authors: Hui Zeng, Jingyuan Xue, Meng Hao, Chen Sun, Bin Ning, Na Zhang
This paper presents CG-Eval, the first comprehensive evaluation of the generation capabilities of large Chinese language models across a wide range of academic disciplines. The models' performance was assessed based on their ability to generate accurate and relevant responses to different types of questions in six disciplines, namely, Science and Engineering, Humanities and Social Sciences, Mathematical Calculations, Medical Practitioner Qualification Examination, Judicial Examination, and Certified Public Accountant Examination. This paper also presents Gscore, a composite index derived from the weighted sum of multiple metrics to measure the quality of model's generation against a reference. The test data and test results can be found at this http URL
Authors: Dawei Gao, Haibin Wang, Yaliang Li, Xiuyu Sun, Yichen Qian, Bolin Ding, Jingren Zhou
Large language models (LLMs) have emerged as a new paradigm for Text-to-SQL task. However, the absence of a systematical benchmark inhibits the development of designing effective, efficient and economic LLM-based Text-to-SQL solutions. To address this challenge, in this paper, we first conduct a systematical and extensive comparison over existing prompt engineering methods, including question representation, example selection and example organization, and with these experimental results, we elaborate their pros and cons. Based on these findings, we propose a new integrated solution, named DAIL-SQL, which refreshes the Spider leaderboard with 86.6% execution accuracy and sets a new bar. To explore the potential of open-source LLM, we investigate them in various scenarios, and further enhance their performance with supervised fine-tuning. Our explorations highlight open-source LLMs' potential in Text-to-SQL, as well as the advantages and disadvantages of the supervised fine-tuning. Additionally, towards an efficient and economic LLM-based Text-to-SQL solution, we emphasize the token efficiency in prompt engineering and compare the prior studies under this metric. We hope that our work provides a deeper understanding of Text-to-SQL with LLMs, and inspires further investigations and broad applications.
Authors: Wei Du, Laksh Advani, Yashmeet Gambhir, Daniel J Perry, Prashant Shiralkar, Zhengzheng Xing, Aaron Colak
Large language models (LLMs) have demonstrated significant capability to generalize across a large number of NLP tasks. For industry applications, it is imperative to assess the performance of the LLM on unlabeled production data from time to time to validate for a real-world setting. Human labeling to assess model error requires considerable expense and time delay. Here we demonstrate that ensemble disagreement scores work well as a proxy for human labeling for language models in zero-shot, few-shot, and fine-tuned settings, per our evaluation on keyphrase extraction (KPE) task. We measure fidelity of the results by comparing to true error measured from human labeled ground truth. We contrast with the alternative of using another LLM as a source of machine labels, or silver labels. Results across various languages and domains show disagreement scores provide a better estimation of model performance with mean average error (MAE) as low as 0.4% and on average 13.8% better than using silver labels.
Authors: Sébastien Ragot
This work proposes to measure the scope of a patent claim as the reciprocal of the self-information contained in this claim. A probability of occurrence of the claim is obtained from a language model and this probability is used to compute the self-information. Grounded in information theory, this approach is based on the assumption that an unlikely concept is more informative than a usual concept, insofar as it is more surprising. In turn, the more surprising the information required to defined the claim, the narrower its scope. Five language models are considered, ranging from simplest models (each word or character is assigned an identical probability) to intermediate models (using average word or character frequencies), to a large language model (GPT2). Interestingly, the scope resulting from the simplest language models is proportional to the reciprocal of the number of words or characters involved in the claim, a metric already used in previous works. Application is made to multiple series of patent claims directed to distinct inventions, where each series consists of claims devised to have a gradually decreasing scope. The performance of the language models is assessed with respect to several ad hoc tests. The more sophisticated the model, the better the results. I.e., the GPT2 probability model outperforms models based on word and character frequencies, which themselves outdo the simplest models based on word or character counts. Still, the character count appears to be a more reliable indicator than the word count.
Authors: Md. Rezaul Karim, Lina Molinas Comet, Md Shajalal, Oya Deniz Beyan, Dietrich Rebholz-Schuhmann, Stefan Decker
Domain experts often rely on most recent knowledge for apprehending and disseminating specific biological processes that help them design strategies for developing prevention and therapeutic decision-making in various disease scenarios. A challenging scenarios for artificial intelligence (AI) is using biomedical data (e.g., texts, imaging, omics, and clinical) to provide diagnosis and treatment recommendations for cancerous conditions.~Data and knowledge about biomedical entities like cancer, drugs, genes, proteins, and their mechanism is spread across structured (knowledge bases (KBs)) and unstructured (e.g., scientific articles) sources. A large-scale knowledge graph (KG) can be constructed by integrating and extracting facts about semantically interrelated entities and relations. Such a KG not only allows exploration and question answering (QA) but also enables domain experts to deduce new knowledge. However, exploring and querying large-scale KGs is tedious for non-domain users due to their lack of understanding of the data assets and semantic technologies. In this paper, we develop a domain KG to leverage cancer-specific biomarker discovery and interactive QA. For this, we constructed a domain ontology called OncoNet Ontology (ONO), which enables semantic reasoning for validating gene-disease (different types of cancer) relations. The KG is further enriched by harmonizing the ONO, metadata, controlled vocabularies, and biomedical concepts from scientific articles by employing BioBERT- and SciBERT-based information extractors. Further, since the biomedical domain is evolving, where new findings often replace old ones, without having access to up-to-date scientific findings, there is a high chance an AI system exhibits concept drift while providing diagnosis and treatment. Therefore, we fine-tune the KG using large language models (LLMs) based on more recent articles and KBs.
Authors: Yi Bin, Wenhao Shi, Yujuan Ding, Yang Yang, See-Kiong Ng
Math word problem (MWP) solving aims to understand the descriptive math problem and calculate the result, for which previous efforts are mostly devoted to upgrade different technical modules. This paper brings a different perspective of \textit{reexamination process} during training by introducing a pseudo-dual task to enhance the MWP solving. We propose a pseudo-dual (PseDual) learning scheme to model such process, which is model-agnostic thus can be adapted to any existing MWP solvers. The pseudo-dual task is specifically defined as filling the numbers in the expression back into the original word problem with numbers masked. To facilitate the effective joint learning of the two tasks, we further design a scheduled fusion strategy for the number infilling task, which smoothly switches the input from the ground-truth math expressions to the predicted ones. Our pseudo-dual learning scheme has been tested and proven effective when being equipped in several representative MWP solvers through empirical studies. \textit{The codes and trained models are available at:} \url{https://github.com/steven640pixel/PsedualMWP}. \end{abstract}
Authors: Chengwei Qin, Chen Chen, Shafiq Joty
Lifelong sequence generation (LSG), a problem in continual learning, aims to continually train a model on a sequence of generation tasks to learn constantly emerging new generation patterns while avoiding the forgetting of previous knowledge. Existing LSG methods mainly focus on maintaining old knowledge while paying little attention to knowledge transfer across tasks. In contrast, humans can better learn new tasks by leveraging previously acquired knowledge from similar tasks. Inspired by the learning paradigm of humans, we propose Dynamic Module Expansion and Adaptation (DMEA), which enables the model to dynamically determine the architecture for acquiring new knowledge based on task correlation and select the most similar previous tasks to facilitate adaptation to new tasks. In addition, as the learning process can easily be biased towards the current task which might cause more severe forgetting of previously learned knowledge, we propose dynamic gradient scaling to balance the learning of the current task and replayed tasks. With extensive experiments, we demonstrate that DMEA can consistently outperform existing methods in different LSG settings.
Authors: Aaquib Syed, Can Rager, Arthur Conmy
Automated interpretability research has recently attracted attention as a potential research direction that could scale explanations of neural network behavior to large models. Existing automated circuit discovery work applies activation patching to identify subnetworks responsible for solving specific tasks (circuits). In this work, we show that a simple method based on attribution patching outperforms all existing methods while requiring just two forward passes and a backward pass. We apply a linear approximation to activation patching to estimate the importance of each edge in the computational subgraph. Using this approximation, we prune the least important edges of the network. We survey the performance and limitations of this method, finding that averaged over all tasks our method has greater AUC from circuit recovery than other methods.
Authors: Guande He, Peng Cui, Jianfei Chen, Wenbo Hu, Jun Zhu
Despite the significant progress made in practical applications of aligned language models (LMs), they tend to be overconfident in output answers compared to the corresponding pre-trained LMs. In this work, we systematically evaluate the impact of the alignment process on logit-based uncertainty calibration of LMs under the multiple-choice setting. We first conduct a thoughtful empirical study on how aligned LMs differ in calibration from their pre-trained counterparts. Experimental results reveal that there are two distinct uncertainties in LMs under the multiple-choice setting, which are responsible for the answer decision and the format preference of the LMs, respectively. Then, we investigate the role of these two uncertainties on aligned LM's calibration through fine-tuning in simple synthetic alignment schemes and conclude that one reason for aligned LMs' overconfidence is the conflation of these two types of uncertainty. Furthermore, we examine the utility of common post-hoc calibration methods for aligned LMs and propose an easy-to-implement and sample-efficient method to calibrate aligned LMs. We hope our findings could provide insights into the design of more reliable alignment processes for LMs.
Authors: Zhengcong Yin, Diya Li, Daniel W. Goldberg
The remarkable success of GPT models across various tasks, including toponymy recognition motivates us to assess the performance of the GPT-3 model in the geocoding address parsing task. To ensure that the evaluation more accurately mirrors performance in real-world scenarios with diverse user input qualities and resolve the pressing need for a 'gold standard' evaluation dataset for geocoding systems, we introduce a benchmark dataset of low-quality address descriptions synthesized based on human input patterns mining from actual input logs of a geocoding system in production. This dataset has 21 different input errors and variations; contains over 239,000 address records that are uniquely selected from streets across all U.S. 50 states and D.C.; and consists of three subsets to be used as training, validation, and testing sets. Building on this, we train and gauge the performance of the GPT-3 model in extracting address components, contrasting its performance with transformer-based and LSTM-based models. The evaluation results indicate that Bidirectional LSTM-CRF model has achieved the best performance over these transformer-based models and GPT-3 model. Transformer-based models demonstrate very comparable results compared to the Bidirectional LSTM-CRF model. The GPT-3 model, though trailing in performance, showcases potential in the address parsing task with few-shot examples, exhibiting room for improvement with additional fine-tuning. We open source the code and data of this presented benchmark so that researchers can utilize it for future model development or extend it to evaluate similar tasks, such as document geocoding.
Authors: Gabriel Sarch, Yue Wu, Michael J. Tarr, Katerina Fragkiadaki
Pre-trained and frozen large language models (LLMs) can effectively map simple scene rearrangement instructions to programs over a robot's visuomotor functions through appropriate few-shot example prompting. To parse open-domain natural language and adapt to a user's idiosyncratic procedures, not known during prompt engineering time, fixed prompts fall short. In this paper, we introduce HELPER, an embodied agent equipped with an external memory of language-program pairs that parses free-form human-robot dialogue into action programs through retrieval-augmented LLM prompting: relevant memories are retrieved based on the current dialogue, instruction, correction, or VLM description, and used as in-context prompt examples for LLM querying. The memory is expanded during deployment to include pairs of user's language and action plans, to assist future inferences and personalize them to the user's language and routines. HELPER sets a new state-of-the-art in the TEACh benchmark in both Execution from Dialog History (EDH) and Trajectory from Dialogue (TfD), with a 1.7x improvement over the previous state-of-the-art for TfD. Our models, code, and video results can be found in our project's website: https://helper-agent-llm.github.io.
Authors: Tao Shi, Xiao Liang, Yaoyuan Liang, Xinyi Tong, Shao-Lun Huang
Emotion recognition in conversations (ERC) is a rapidly evolving task within the natural language processing community, which aims to detect the emotions expressed by speakers during a conversation. Recently, a growing number of ERC methods have focused on leveraging supervised contrastive learning (SCL) to enhance the robustness and generalizability of learned features. However, current SCL-based approaches in ERC are impeded by the constraint of large batch sizes and the lack of compatibility with most existing ERC models. To address these challenges, we propose an efficient and model-agnostic SCL framework named Supervised Sample-Label Contrastive Learning with Soft-HGR Maximal Correlation (SSLCL), which eliminates the need for a large batch size and can be seamlessly integrated with existing ERC models without introducing any model-specific assumptions. Specifically, we introduce a novel perspective on utilizing label representations by projecting discrete labels into dense embeddings through a shallow multilayer perceptron, and formulate the training objective to maximize the similarity between sample features and their corresponding ground-truth label embeddings, while minimizing the similarity between sample features and label embeddings of disparate classes. Moreover, we innovatively adopt the Soft-HGR maximal correlation as a measure of similarity between sample features and label embeddings, leading to significant performance improvements over conventional similarity measures. Additionally, multimodal cues of utterances are effectively leveraged by SSLCL as data augmentations to boost model performances. Extensive experiments on two ERC benchmark datasets, IEMOCAP and MELD, demonstrate the compatibility and superiority of our proposed SSLCL framework compared to existing state-of-the-art SCL methods. Our code is available at \url{https://github.com/TaoShi1998/SSLCL}.
Authors: Luyang Fang, Gyeong-Geon Lee, Xiaoming Zhai
Machine learning-based automatic scoring can be challenging if students' responses are unbalanced across scoring categories, as it introduces uncertainty in the machine training process. To meet this challenge, we introduce a novel text data augmentation framework using GPT-4, a generative large language model, specifically tailored for unbalanced datasets in automatic scoring. Our experimental dataset comprised student-written responses to two science items. We crafted prompts for GPT-4 to generate responses resembling student-written answers, particularly for the minority scoring classes, to augment the data. We then finetuned DistillBERT for automatic scoring based on the augmented and original datasets. Model performance was assessed using accuracy, precision, recall, and F1 score. We incorporate varied amounts of augmented data to examine scoring performance, and our findings revealed remarkedly improved model performance. The average maximum increase observed across two items is: 3.5% for accuracy, 30.6% for precision, 21.1% for recall, and 24.2% for F1 score. Notably, using just 5% of the augmented data led to substantial improvements: 2.6%, 29.2%, 15.1%, and 19.6%. Interestingly, the extent of improvement varied depending on specific datasets. Moreover, we found that a varying amount of augmented data (5%-40%) was needed to obtain a stable improvement. We also compare models trained with GPT-4 augmented data and those trained with additional student-written responses. The findings indicate that former ones match or even exceed the performance of the latter. Specifically, there is an average difference of 1.7%, 1.9%, 11.0%, and 7.8% for four metrics separately. This research underscores the potential and effectiveness of data augmentation techniques utilizing GPT-4 in addressing unbalanced datasets within automated assessment.
Authors: Ahmed Sabir, Lluís Padró
In this paper, we investigate the impact of objects on gender bias in image captioning systems. Our results show that only gender-specific objects have a strong gender bias (e.g., women-lipstick). In addition, we propose a visual semantic-based gender score that measures the degree of bias and can be used as a plug-in for any image captioning system. Our experiments demonstrate the utility of the gender score, since we observe that our score can measure the bias relation between a caption and its related gender; therefore, our score can be used as an additional metric to the existing Object Gender Co-Occ approach. Code and data are publicly available at \url{https://github.com/ahmedssabir/GenderScore}.
Authors: Samuel Belkadi, Nicolo Micheletti, Lifeng Han, Warren Del-Pinto, Goran Nenadic
Access to real-world medication prescriptions is essential for medical research and healthcare quality improvement. However, access to real medication prescriptions is often limited due to the sensitive nature of the information expressed. Additionally, manually labelling these instructions for training and fine-tuning Natural Language Processing (NLP) models can be tedious and expensive. We introduce a novel task-specific model architecture, Label-To-Text-Transformer (\textbf{LT3}), tailored to generate synthetic medication prescriptions based on provided labels, such as a vocabulary list of medications and their attributes. LT3 is trained on a set of around 2K lines of medication prescriptions extracted from the MIMIC-III database, allowing the model to produce valuable synthetic medication prescriptions. We evaluate LT3's performance by contrasting it with a state-of-the-art Pre-trained Language Model (PLM), T5, analysing the quality and diversity of generated texts. We deploy the generated synthetic data to train the SpacyNER model for the Named Entity Recognition (NER) task over the n2c2-2018 dataset. The experiments show that the model trained on synthetic data can achieve a 96-98\% F1 score at Label Recognition on Drug, Frequency, Route, Strength, and Form. LT3 codes and data will be shared at \url{https://github.com/HECTA-UoM/Label-To-Text-Transformer}
Authors: Khushi Bhardwaj, Raj Sanjay Shah, Sashank Varma
Pre-trained Large Language Models (LLMs) have shown success in a diverse set of language inference and understanding tasks. The pre-training stage of LLMs looks at a large corpus of raw textual data. The BabyLM shared task compares LLM pre-training to human language acquisition, where the number of tokens seen by 13-year-old kids is magnitudes smaller than the number of tokens seen by LLMs. In this work, we pre-train and evaluate LLMs on their ability to learn contextual word representations using roughly the same number of tokens as seen by children. We provide a strong set of baselines; with different architectures, evaluation of changes in performance across epochs, and reported pre-training metrics for the strict small and strict tracks of the task. We also try to loosely replicate the RoBERTa baseline given by the task organizers to observe the training robustness to hyperparameter selection and replicability. We provide the submission details to the strict and strict-small tracks in this report.
Authors: Vatsal Venkatkrishna, Durga Shree Nagabushanam, Emmanuel Iko-Ojo Simon, Melina Vidoni
Documentation debt hinders the effective utilization of open-source software. Although code summarization tools have been helpful for developers, most would prefer a detailed account of each parameter in a function rather than a high-level summary. However, generating such a summary is too intricate for a single generative model to produce reliably due to the lack of high-quality training data. Thus, we propose a multi-step approach that combines multiple task-specific models, each adept at producing a specific section of a docstring. The combination of these models ensures the inclusion of each section in the final docstring. We compared the results from our approach with existing generative models using both automatic metrics and a human-centred evaluation with 17 participating developers, which proves the superiority of our approach over existing methods.
Authors: Joshua Clymer, Garrett Baker, Rohan Subramani, Sam Wang
As AI systems become more intelligent and their behavior becomes more challenging to assess, they may learn to game the flaws of human feedback instead of genuinely striving to follow instructions; however, this risk can be mitigated by controlling how LLMs generalize human feedback to situations where it is unreliable. To better understand how reward models generalize, we craft 69 distribution shifts spanning 8 categories. We find that reward models do not learn to evaluate `instruction-following' by default and instead favor personas that resemble internet text. Techniques for interpreting reward models' internal representations achieve better generalization than standard fine-tuning, but still frequently fail to distinguish instruction-following from conflated behaviors. We consolidate the 15 most challenging distribution shifts into the GENeralization analogIES (GENIES) benchmark, which we hope will enable progress toward controlling reward model generalization.
Authors: Gilad Deutch, Nadav Magar, Tomer Bar Natan, Guy Dar
In-context learning (ICL) has shown impressive results in few-shot learning tasks, yet its underlying mechanism is still not fully understood. Recent works suggest that ICL can be thought of as a gradient descent (GD) based optimization process. While promising, these results mainly focus on simplified settings of ICL and provide only a preliminary evaluation of the similarities between the two methods. In this work, we revisit the comparison between ICL and GD-based finetuning and study what properties of ICL an equivalent process must follow. We highlight a major difference in the flow of information between ICL and standard finetuning. Namely, ICL can only rely on information from lower layers at every point, while finetuning depends on loss gradients from deeper layers. We refer to this discrepancy as Layer Causality and show that a layer causal variant of the finetuning process aligns with ICL on par with vanilla finetuning and is even better in most cases across relevant metrics. To the best of our knowledge, this is the first work to discuss this discrepancy explicitly and suggest a solution that tackles this problem with minimal changes.
Authors: Ziyin Zhang, Chaoyu Chen, Bingchang Liu, Cong Liao, Zi Gong, Hang Yu, Jianguo Li, Rui Wang
In this work we systematically review the recent advancements in code processing with language models, covering 50+ models, 30+ evaluation tasks, 150+ datasets, and 550 related works. We break down code processing models into general language models represented by the GPT family and specialized models that are specifically pretrained on code, often with tailored objectives. We discuss the relations and differences between these models, and highlight the historical transition of code modeling from statistical models and RNNs to pretrained Transformers and LLMs, which is exactly the same course that had been taken by NLP. We also discuss code-specific features such as AST, CFG, and unit tests, along with their application in training code language models, and identify key challenges and potential future directions in this domain. We keep the survey open and updated on GitHub repository at https://github.com/codefuse-ai/Awesome-Code-LLM.
Authors: Ziyi Ye, Qingyao Ai, Yiqun Liu, Min Zhang, Christina Lioma, Tuukka Ruotsalo
Generating human language through non-invasive brain-computer interfaces (BCIs) has the potential to unlock many applications, such as serving disabled patients and improving communication. Currently, however, generating language via BCIs has been previously successful only within a classification setup for selecting pre-generated sentence continuation candidates with the most likely cortical semantic representation. Inspired by recent research that revealed associations between the brain and the large computational language models, we propose a generative language BCI that utilizes the capacity of a large language model (LLM) jointly with a semantic brain decoder to directly generate language from functional magnetic resonance imaging (fMRI) input. The proposed model can generate coherent language sequences aligned with the semantic content of visual or auditory language stimuli perceived, without prior knowledge of any pre-generated candidates. We compare the language generated from the presented model with a random control, pre-generated language selection approach, and a standard LLM, which generates common coherent text solely based on the next word likelihood according to statistical language training data. The proposed model is found to generate language that is more aligned with semantic stimulus in response to which brain input is sampled. Our findings demonstrate the potential and feasibility of employing BCIs in direct language generation.
Authors: Ilaria Manco, Benno Weck, SeungHeon Doh, Minz Won, Yixiao Zhang, Dmitry Bodganov, Yusong Wu, Ke Chen, Philip Tovstogan, Emmanouil Benetos, Elio Quinton, György Fazekas, Juhan Nam
We introduce the Song Describer dataset (SDD), a new crowdsourced corpus of high-quality audio-caption pairs, designed for the evaluation of music-and-language models. The dataset consists of 1.1k human-written natural language descriptions of 706 music recordings, all publicly accessible and released under Creative Common licenses. To showcase the use of our dataset, we benchmark popular models on three key music-and-language tasks (music captioning, text-to-music generation and music-language retrieval). Our experiments highlight the importance of cross-dataset evaluation and offer insights into how researchers can use SDD to gain a broader understanding of model performance.
Authors: Vasilii A. Gromov, Nikita S. Borodin, Asel S. Yerbolova
The present paper introduces a novel object of study - a language fractal structure. We hypothesize that a set of embeddings of all $n$-grams of a natural language constitutes a representative sample of this fractal set. (We use the term Hailonakea to refer to the sum total of all language fractal structures, over all $n$). The paper estimates intrinsic (genuine) dimensions of language fractal structures for the Russian and English languages. To this end, we employ methods based on (1) topological data analysis and (2) a minimum spanning tree of a data graph for a cloud of points considered (Steele theorem). For both languages, for all $n$, the intrinsic dimensions appear to be non-integer values (typical for fractal sets), close to 9 for both of the Russian and English language.
Authors: Hamish Ivison, Yizhong Wang, Valentina Pyatkin, Nathan Lambert, Matthew Peters, Pradeep Dasigi, Joel Jang, David Wadden, Noah A. Smith, Iz Beltagy, Hannaneh Hajishirzi
Since the release of T\"ULU [Wang et al., 2023b], open resources for instruction tuning have developed quickly, from better base models to new finetuning techniques. We test and incorporate a number of these advances into T\"ULU, resulting in T\"ULU 2, a suite of improved T\"ULU models for advancing the understanding and best practices of adapting pretrained language models to downstream tasks and user preferences. Concretely, we release: (1) T\"ULU-V2-mix, an improved collection of high-quality instruction datasets; (2) T\"ULU 2, LLAMA-2 models finetuned on the V2 mixture; (3) T\"ULU 2+DPO, T\"ULU 2 models trained with direct preference optimization (DPO), including the largest DPO-trained model to date (T\"ULU 2+DPO 70B); (4) CODE T\"ULU 2, CODE LLAMA models finetuned on our V2 mix that outperform CODE LLAMA and its instruction-tuned variant, CODE LLAMA-Instruct. Our evaluation from multiple perspectives shows that the T\"ULU 2 suite achieves state-of-the-art performance among open models and matches or exceeds the performance of GPT-3.5-turbo-0301 on several benchmarks. We release all the checkpoints, data, training and evaluation code to facilitate future open efforts on adapting large language models.