In the era of large language models, Mixture-of-Experts (MoE) is a promising architecture for managing computational costs when scaling up model parameters. However, conventional MoE architectures like GShard, which activate the top-$K$ out of $N$ experts, face challenges in ensuring expert specialization, i.e. each expert acquires non-overlapping and focused knowledge. In response, we propose the DeepSeekMoE architecture towards ultimate expert specialization. It involves two principal strategies: (1) finely segmenting the experts into $mN$ ones and activating $mK$ from them, allowing for a more flexible combination of activated experts; (2) isolating $K_s$ experts as shared ones, aiming at capturing common knowledge and mitigating redundancy in routed experts. Starting from a modest scale with 2B parameters, we demonstrate that DeepSeekMoE 2B achieves comparable performance with GShard 2.9B, which has 1.5 times the expert parameters and computation. In addition, DeepSeekMoE 2B nearly approaches the performance of its dense counterpart with the same number of total parameters, which set the upper bound of MoE models. Subsequently, we scale up DeepSeekMoE to 16B parameters and show that it achieves comparable performance with LLaMA2 7B, with only about 40% of computations. Further, our preliminary efforts to scale up DeepSeekMoE to 145B parameters consistently validate its substantial advantages over the GShard architecture, and show its performance comparable with DeepSeek 67B, using only 28.5% (maybe even 18.2%) of computations.
Pretrained language models (PLMs) have become remarkably adept at task and language generalization. Nonetheless, they often fail when faced with unseen languages. In this work, we present LinguAlchemy, a regularization method that incorporates various linguistic information covering typological, geographical, and phylogenetic features to align PLMs representation to the corresponding linguistic information on each language. Our LinguAlchemy significantly improves the performance of mBERT and XLM-R on low-resource languages in multiple downstream tasks such as intent classification, news classification, and semantic relatedness compared to fully finetuned models and displaying a high degree of unseen language generalization. We further introduce AlchemyScale and AlchemyTune, extension of LinguAlchemy which adjusts the linguistic regularization weights automatically, alleviating the need for hyperparameter search.
The recent advance in vision-language models is largely attributed to the abundance of image-text data. We aim to replicate this success for video-language models, but there simply is not enough human-curated video-text data available. We thus resort to fine-tuning a video-language model from a strong image-language baseline with synthesized instructional data. The resulting video model by video-instruction-tuning (VIIT) is then used to auto-label millions of videos to generate high-quality captions. We show the adapted video-language model performs well on a wide range of video-language benchmarks. For instance, it surpasses the best prior result on open-ended NExT-QA by 2.8%. Besides, our model generates detailed descriptions for previously unseen videos, which provide better textual supervision than existing methods. Experiments show that a video-language dual-encoder model contrastively trained on these auto-generated captions is 3.8% better than the strongest baseline that also leverages vision-language models. Our best model outperforms state-of-the-art methods on MSR-VTT zero-shot text-to-video retrieval by 6%. As a side product, we generate the largest video caption dataset to date.
Multi-modal large language models have demonstrated impressive performance across various tasks in different modalities. However, existing multi-modal models primarily emphasize capturing global information within each modality while neglecting the importance of perceiving local information across modalities. Consequently, these models lack the ability to effectively understand the fine-grained details of input data, limiting their performance in tasks that require a more nuanced understanding. To address this limitation, there is a compelling need to develop models that enable fine-grained understanding across multiple modalities, thereby enhancing their applicability to a wide range of tasks. In this paper, we propose GroundingGPT, a language enhanced multi-modal grounding model. Beyond capturing global information like other multi-modal models, our proposed model excels at tasks demanding a detailed understanding of local information within the input. It demonstrates precise identification and localization of specific regions in images or moments in videos. To achieve this objective, we design a diversified dataset construction pipeline, resulting in a multi-modal, multi-granularity dataset for model training. The code, dataset, and demo of our model can be found at https: //github.com/lzw-lzw/GroundingGPT.
With large training datasets and massive amounts of computing sources, large language models (LLMs) achieve remarkable performance in comprehensive and generative ability. Based on those powerful LLMs, the model fine-tuned with domain-specific datasets posseses more specialized knowledge and thus is more practical like medical LLMs. However, the existing fine-tuned medical LLMs are limited to general medical knowledge with English language. For disease-specific problems, the model's response is inaccurate and sometimes even completely irrelevant, especially when using a language other than English. In this work, we focus on the particular disease of Epilepsy with Japanese language and introduce a customized LLM termed as EpilepsyLLM. Our model is trained from the pre-trained LLM by fine-tuning technique using datasets from the epilepsy domain. The datasets contain knowledge of basic information about disease, common treatment methods and drugs, and important notes in life and work. The experimental results demonstrate that EpilepsyLLM can provide more reliable and specialized medical knowledge responses.
Researchers have investigated the potential of leveraging pre-trained language models, such as CodeBERT, to enhance source code-related tasks. Previous methodologies have relied on CodeBERT's '[CLS]' token as the embedding representation of input sequences for task performance, necessitating additional neural network layers to enhance feature representation, which in turn increases computational expenses. These approaches have also failed to fully leverage the comprehensive knowledge inherent within the source code and its associated text, potentially limiting classification efficacy. We propose CodeClassPrompt, a text classification technique that harnesses prompt learning to extract rich knowledge associated with input sequences from pre-trained models, thereby eliminating the need for additional layers and lowering computational costs. By applying an attention mechanism, we synthesize multi-layered knowledge into task-specific features, enhancing classification accuracy. Our comprehensive experimentation across four distinct source code-related tasks reveals that CodeClassPrompt achieves competitive performance while significantly reducing computational overhead.
The objective of this study is to improve automated feedback tools designed for English Language Learners (ELLs) through the utilization of data science techniques encompassing machine learning, natural language processing, and educational data analytics. Automated essay scoring (AES) research has made strides in evaluating written essays, but it often overlooks the specific needs of English Language Learners (ELLs) in language development. This study explores the application of BERT-related techniques to enhance the assessment of ELLs' writing proficiency within AES. To address the specific needs of ELLs, we propose the use of DeBERTa, a state-of-the-art neural language model, for improving automated feedback tools. DeBERTa, pretrained on large text corpora using self-supervised learning, learns universal language representations adaptable to various natural language understanding tasks. The model incorporates several innovative techniques, including adversarial training through Adversarial Weights Perturbation (AWP) and Metric-specific AttentionPooling (6 kinds of AP) for each label in the competition. The primary focus of this research is to investigate the impact of hyperparameters, particularly the adversarial learning rate, on the performance of the model. By fine-tuning the hyperparameter tuning process, including the influence of 6AP and AWP, the resulting models can provide more accurate evaluations of language proficiency and support tailored learning tasks for ELLs. This work has the potential to significantly benefit ELLs by improving their English language proficiency and facilitating their educational journey.
Language models pre-trained on web-scale corpora demonstrate impressive capabilities on diverse downstream tasks. However, there is increasing concern whether such capabilities might arise from evaluation datasets being included in the pre-training corpus -- a phenomenon known as \textit{data contamination} -- in a manner that artificially increases performance. There has been little understanding of how this potential contamination might influence LMs' performance on downstream tasks. In this paper, we explore the impact of data contamination at the pre-training stage by pre-training a series of GPT-2 models \textit{from scratch}. We highlight the effect of both text contamination (\textit{i.e.}\ input text of the evaluation samples) and ground-truth contamination (\textit{i.e.}\ the prompts asked on the input and the desired outputs) from evaluation data. We also investigate the effects of repeating contamination for various downstream tasks. Additionally, we examine the prevailing n-gram-based definitions of contamination within current LLM reports, pinpointing their limitations and inadequacy. Our findings offer new insights into data contamination's effects on language model capabilities and underscore the need for independent, comprehensive contamination assessments in LLM studies.
While state-of-the-art language models have achieved impressive results, they remain susceptible to inference-time adversarial attacks, such as adversarial prompts generated by red teams arXiv:2209.07858. One approach proposed to improve the general quality of language model generations is multi-agent debate, where language models self-evaluate through discussion and feedback arXiv:2305.14325. We implement multi-agent debate between current state-of-the-art language models and evaluate models' susceptibility to red team attacks in both single- and multi-agent settings. We find that multi-agent debate can reduce model toxicity when jailbroken or less capable models are forced to debate with non-jailbroken or more capable models. We also find marginal improvements through the general usage of multi-agent interactions. We further perform adversarial prompt content classification via embedding clustering, and analyze the susceptibility of different models to different types of attack topics.
Question generation (QG) is a natural language processing task with an abundance of potential benefits and use cases in the educational domain. In order for this potential to be realized, QG systems must be designed and validated with pedagogical needs in mind. However, little research has assessed or designed QG approaches with the input from real teachers or students. This paper applies a large language model-based QG approach where questions are generated with learning goals derived from Bloom's taxonomy. The automatically generated questions are used in multiple experiments designed to assess how teachers use them in practice. The results demonstrate that teachers prefer to write quizzes with automatically generated questions, and that such quizzes have no loss in quality compared to handwritten versions. Further, several metrics indicate that automatically generated questions can even improve the quality of the quizzes created, showing the promise for large scale use of QG in the classroom setting.
Large language models (LLMs) have strong capabilities in solving diverse natural language processing tasks. However, the safety and security issues of LLM systems have become the major obstacle to their widespread application. Many studies have extensively investigated risks in LLM systems and developed the corresponding mitigation strategies. Leading-edge enterprises such as OpenAI, Google, Meta, and Anthropic have also made lots of efforts on responsible LLMs. Therefore, there is a growing need to organize the existing studies and establish comprehensive taxonomies for the community. In this paper, we delve into four essential modules of an LLM system, including an input module for receiving prompts, a language model trained on extensive corpora, a toolchain module for development and deployment, and an output module for exporting LLM-generated content. Based on this, we propose a comprehensive taxonomy, which systematically analyzes potential risks associated with each module of an LLM system and discusses the corresponding mitigation strategies. Furthermore, we review prevalent benchmarks, aiming to facilitate the risk assessment of LLM systems. We hope that this paper can help LLM participants embrace a systematic perspective to build their responsible LLM systems.
At the heart of medicine lies the physician-patient dialogue, where skillful history-taking paves the way for accurate diagnosis, effective management, and enduring trust. Artificial Intelligence (AI) systems capable of diagnostic dialogue could increase accessibility, consistency, and quality of care. However, approximating clinicians' expertise is an outstanding grand challenge. Here, we introduce AMIE (Articulate Medical Intelligence Explorer), a Large Language Model (LLM) based AI system optimized for diagnostic dialogue. AMIE uses a novel self-play based simulated environment with automated feedback mechanisms for scaling learning across diverse disease conditions, specialties, and contexts. We designed a framework for evaluating clinically-meaningful axes of performance including history-taking, diagnostic accuracy, management reasoning, communication skills, and empathy. We compared AMIE's performance to that of primary care physicians (PCPs) in a randomized, double-blind crossover study of text-based consultations with validated patient actors in the style of an Objective Structured Clinical Examination (OSCE). The study included 149 case scenarios from clinical providers in Canada, the UK, and India, 20 PCPs for comparison with AMIE, and evaluations by specialist physicians and patient actors. AMIE demonstrated greater diagnostic accuracy and superior performance on 28 of 32 axes according to specialist physicians and 24 of 26 axes according to patient actors. Our research has several limitations and should be interpreted with appropriate caution. Clinicians were limited to unfamiliar synchronous text-chat which permits large-scale LLM-patient interactions but is not representative of usual clinical practice. While further research is required before AMIE could be translated to real-world settings, the results represent a milestone towards conversational diagnostic AI.
This paper proposes a classification framework aimed at identifying correlations between job ad requirements and transversal skill sets, with a focus on predicting the necessary skills for individual job descriptions using a deep learning model. The approach involves data collection, preprocessing, and labeling using ESCO (European Skills, Competences, and Occupations) taxonomy. Hierarchical classification and multi-label strategies are used for skill identification, while augmentation techniques address data imbalance, enhancing model robustness. A comparison between results obtained with English-specific and multi-language sentence embedding models reveals close accuracy. The experimental case studies detail neural network configurations, hyperparameters, and cross-validation results, highlighting the efficacy of the hierarchical approach and the suitability of the multi-language model for the diverse European job market. Thus, a new approach is proposed for the hierarchical classification of transversal skills from job ads.
Large Language Models have shown exceptional generative abilities in various natural language and generation tasks. However, possible anthropomorphization and leniency towards failure cases have propelled discussions on emergent abilities of Large Language Models especially on Theory of Mind (ToM) abilities in Large Language Models. While several false-belief tests exists to verify the ability to infer and maintain mental models of another entity, we study a special application of ToM abilities that has higher stakes and possibly irreversible consequences : Human Robot Interaction. In this work, we explore the task of Perceived Behavior Recognition, where a robot employs a Large Language Model (LLM) to assess the robot's generated behavior in a manner similar to human observer. We focus on four behavior types, namely - explicable, legible, predictable, and obfuscatory behavior which have been extensively used to synthesize interpretable robot behaviors. The LLMs goal is, therefore to be a human proxy to the agent, and to answer how a certain agent behavior would be perceived by the human in the loop, for example "Given a robot's behavior X, would the human observer find it explicable?". We conduct a human subject study to verify that the users are able to correctly answer such a question in the curated situations (robot setting and plan) across five domains. A first analysis of the belief test yields extremely positive results inflating ones expectations of LLMs possessing ToM abilities. We then propose and perform a suite of perturbation tests which breaks this illusion, i.e. Inconsistent Belief, Uninformative Context and Conviction Test. We conclude that, the high score of LLMs on vanilla prompts showcases its potential use in HRI settings, however to possess ToM demands invariance to trivial or irrelevant perturbations in the context which LLMs lack.
Recent advances in natural language processing enable more intelligent ways to support knowledge sharing in factories. In manufacturing, operating production lines has become increasingly knowledge-intensive, putting strain on a factory's capacity to train and support new operators. This paper introduces a Large Language Model (LLM)-based system designed to retrieve information from the extensive knowledge contained in factory documentation and knowledge shared by expert operators. The system aims to efficiently answer queries from operators and facilitate the sharing of new knowledge. We conducted a user study at a factory to assess its potential impact and adoption, eliciting several perceived benefits, namely, enabling quicker information retrieval and more efficient resolution of issues. However, the study also highlighted a preference for learning from a human expert when such an option is available. Furthermore, we benchmarked several commercial and open-sourced LLMs for this system. The current state-of-the-art model, GPT-4, consistently outperformed its counterparts, with open-source models trailing closely, presenting an attractive option given their data privacy and customization benefits. In summary, this work offers preliminary insights and a system design for factories considering using LLM tools for knowledge management.
One of the most important challenges in text generation systems is to produce outputs that are not only correct but also diverse. Recently, Minimum Bayes-Risk (MBR) decoding has gained prominence for generating sentences of the highest quality among the decoding algorithms. However, existing algorithms proposed for generating diverse outputs are predominantly based on beam search or random sampling, thus their output quality is capped by these underlying methods. In this paper, we investigate an alternative approach -- we develop diversity-promoting decoding algorithms by enforcing diversity objectives to MBR decoding. We propose two variants of MBR, Diverse MBR (DMBR) and $k$-medoids MBR (KMBR), methods to generate a set of sentences with high quality and diversity. We evaluate DMBR and KMBR on a variety of directed text generation tasks using encoder-decoder models and a large language model with prompting. The experimental results show that the proposed method achieves a better trade-off than the diverse beam search and sampling algorithms.
Few-shot Learning aims to learn and distinguish new categories with a very limited number of available images, presenting a significant challenge in the realm of deep learning. Recent researchers have sought to leverage the additional textual or linguistic information of these rare categories with a pre-trained language model to facilitate learning, thus partially alleviating the problem of insufficient supervision signals. However, the full potential of the textual information and pre-trained language model have been underestimated in the few-shot learning till now, resulting in limited performance enhancements. To address this, we propose a simple but effective framework for few-shot learning tasks, specifically designed to exploit the textual information and language model. In more detail, we explicitly exploit the zero-shot capability of the pre-trained language model with the learnable prompt. And we just add the visual feature with the textual feature for inference directly without the intricate designed fusion modules in previous works. Additionally, we apply the self-ensemble and distillation to further enhance these components. Our extensive experiments conducted across four widely used few-shot datasets demonstrate that our simple framework achieves impressive results. Particularly noteworthy is its outstanding performance in the 1-shot learning task, surpassing state-of-the-art methods by an average of 3.0\% in classification accuracy. \footnote{We will make the source codes of the proposed framework publicly available upon acceptance. }.
We investigate large language model performance across five orders of magnitude of compute scaling in eleven recent model architectures. We show that average benchmark performance, aggregating over many individual tasks and evaluations as in the commonly-used BIG-Bench dataset, is decently predictable as a function of training compute scale. Specifically, when extrapolating BIG-Bench Hard performance across one order of magnitude in compute, we observe average absolute errors of 6 percentage points (pp). By contrast, extrapolation for individual BIG-Bench tasks across an order of magnitude in compute yields higher average errors of 18pp. Nonetheless, individual task performance remains significantly more predictable than chance. Overall, our work suggests compute scaling provides a promising basis to forecast AI capabilities in diverse benchmarks, though predicting performance in specific tasks poses challenges.
People are relying on AI agents to assist them with various tasks. The human must know when to rely on the agent, collaborate with the agent, or ignore its suggestions. In this work, we propose to learn rules, grounded in data regions and described in natural language, that illustrate how the human should collaborate with the AI. Our novel region discovery algorithm finds local regions in the data as neighborhoods in an embedding space where prior human behavior should be corrected. Each region is then described using a large language model in an iterative and contrastive procedure. We then teach these rules to the human via an onboarding stage. Through user studies on object detection and question-answering tasks, we show that our method can lead to more accurate human-AI teams. We also evaluate our region discovery and description algorithms separately.
In code search, the Generation-Augmented Retrieval (GAR) framework, which generates exemplar code snippets to augment queries, has emerged as a promising strategy to address the principal challenge of modality misalignment between code snippets and natural language queries, particularly with the demonstrated code generation capabilities of Large Language Models (LLMs). Nevertheless, our preliminary investigations indicate that the improvements conferred by such an LLM-augmented framework are somewhat constrained. This limitation could potentially be ascribed to the fact that the generated codes, albeit functionally accurate, frequently display a pronounced stylistic deviation from the ground truth code in the codebase. In this paper, we extend the foundational GAR framework and propose a simple yet effective method that additionally Rewrites the Code (ReCo) within the codebase for style normalization. Experimental results demonstrate that ReCo significantly boosts retrieval accuracy across sparse (up to 35.7%), zero-shot dense (up to 27.6%), and fine-tuned dense (up to 23.6%) retrieval settings in diverse search scenarios. To further elucidate the advantages of ReCo and stimulate research in code style normalization, we introduce Code Style Similarity, the first metric tailored to quantify stylistic similarities in code. Notably, our empirical findings reveal the inadequacy of existing metrics in capturing stylistic nuances. The source code and data are available at \url{https://github.com/Alex-HaochenLi/ReCo}.