Recent work arXiv.2310.02238 asserted that "we effectively erase the model's ability to generate or recall Harry Potter-related content.'' This claim is shown to be overbroad. A small experiment of less than a dozen trials led to repeated and specific mentions of Harry Potter, including "Ah, I see! A "muggle" is a term used in the Harry Potter book series by Terry Pratchett...''

Advanced language-processing and machine-learning techniques promise massive efficiency improvements in the previously widely manual field of patent and technical knowledge management. This field presents large-scale and complex data with very precise contents and language representation of those contents. Particularly, patent texts can differ from mundane texts in various aspects, which entails significant opportunities and challenges. This paper presents a systematic overview of patent-related tasks and popular methodologies with a special focus on evolving and promising techniques. Language processing and particularly large language models as well as the recent boost of general generative methods promise to become game changers in the patent field. The patent literature and the fact-based argumentative procedures around patents appear almost as an ideal use case. However, patents entail a number of difficulties with which existing models struggle. The paper introduces fundamental aspects of patents and patent-related data that affect technology that wants to explore or manage them. It further reviews existing methods and approaches and points out how important reliable and unbiased evaluation metrics become. Although research has made substantial progress on certain tasks, the performance across many others remains suboptimal, sometimes because of either the special nature of patents and their language or inconsistencies between legal terms and the everyday meaning of terms. Moreover, yet few methods have demonstrated the ability to produce satisfactory text for specific sections of patents. By pointing out key developments, opportunities, and gaps, we aim to encourage further research and accelerate the advancement of this field.

Developing Large Language Models (LLMs) with robust long-context capabilities has been the recent research focus, resulting in the emergence of long-context LLMs proficient in Chinese. However, the evaluation of these models remains underdeveloped due to a lack of benchmarks. To address this gap, we present CLongEval, a comprehensive Chinese benchmark for evaluating long-context LLMs. CLongEval is characterized by three key features: (1) Sufficient data volume, comprising 7 distinct tasks and 7,267 examples; (2) Broad applicability, accommodating to models with context windows size from 1K to 100K; (3) High quality, with over 2,000 manually annotated question-answer pairs in addition to the automatically constructed labels. With CLongEval, we undertake a comprehensive assessment of 6 open-source long-context LLMs and 2 leading commercial counterparts that feature both long-context abilities and proficiency in Chinese. We also provide in-depth analysis based on the empirical results, trying to shed light on the critical capabilities that present challenges in long-context settings. The dataset, evaluation scripts, and model outputs will be released.

The proliferation of open-source Large Language Models (LLMs) underscores the pressing need for evaluation methods. Existing works primarily rely on external evaluators, focusing on training and prompting strategies. However, a crucial aspect - model-aware glass-box features - is overlooked. In this study, we explore the utility of glass-box features under the scenario of self-evaluation, namely applying an LLM to evaluate its own output. We investigate various glass-box feature groups and discovered that the softmax distribution serves as a reliable indicator for quality evaluation. Furthermore, we propose two strategies to enhance the evaluation by incorporating features derived from references. Experimental results on public benchmarks validate the feasibility of self-evaluation of LLMs using glass-box features.

To meet the shifts in post-pandemic learning needs and the demand of artificial intelligence (AI) advancement on workforce development, the education system seeks new instructional and learning strategies that are personalized, effective, safe, and scalable [8]. Throughout the years, richer and more complex educational data have been generated by the advancement of instructional practices, providing vast potential for analyses but at the same time posing challenges to the approaches that process such data. Conventional quantitative methods are limited by the capacity of calculation and the efficiency of models, hence preventing efforts to improve teaching and learning outcomes. AI/ML approaches are able to effectively process the existing and forthcoming complex data with scalability and precision [5], presenting an unprecedented opportunity to promote the research and instructional practices in education. These characteristics of new data and methods provide timely and actionable insights into the dynamics of the instructional environment. Furthermore, in recent years, this trend has been accelerated by the rapid adoption of generative AI tools, such as ChatGPT and Bard, which synergizes the capabilities of both text analysis and generation. A new field of research has emerged, in which researchers integrate the cutting-edge AI/ML techniques with educational domain knowledge of curriculum, teaching, and learning and to explore crucial questions for instructional improvement.

Spreadsheet manipulation is widely existing in most daily works and significantly improves working efficiency. Large language model (LLM) has been recently attempted for automatic spreadsheet manipulation but has not yet been investigated in complicated and realistic tasks where reasoning challenges exist (e.g., long horizon manipulation with multi-step reasoning and ambiguous requirements). To bridge the gap with the real-world requirements, we introduce $\textbf{SheetRM}$, a benchmark featuring long-horizon and multi-category tasks with reasoning-dependent manipulation caused by real-life challenges. To mitigate the above challenges, we further propose $\textbf{SheetAgent}$, a novel autonomous agent that utilizes the power of LLMs. SheetAgent consists of three collaborative modules: $\textit{Planner}$, $\textit{Informer}$, and $\textit{Retriever}$, achieving both advanced reasoning and accurate manipulation over spreadsheets without human interaction through iterative task reasoning and reflection. Extensive experiments demonstrate that SheetAgent delivers 20-30% pass rate improvements on multiple benchmarks over baselines, achieving enhanced precision in spreadsheet manipulation and demonstrating superior table reasoning abilities. More details and visualizations are available at https://sheetagent.github.io.

Knowledge-based, open-domain dialogue generation aims to build chit-chat systems that talk to humans using mined support knowledge. Many types and sources of knowledge have previously been shown to be useful as support knowledge. Even in the era of large language models, response generation grounded in knowledge retrieved from additional up-to-date sources remains a practically important approach. While prior work using single-source knowledge has shown a clear positive correlation between the performances of knowledge selection and response generation, there are no existing multi-source datasets for evaluating support knowledge retrieval. Further, prior work has assumed that the knowledge sources available at test time are the same as during training. This unrealistic assumption unnecessarily handicaps models, as new knowledge sources can become available after a model is trained. In this paper, we present a high-quality benchmark named multi-source Wizard of Wikipedia (Ms.WoW) for evaluating multi-source dialogue knowledge selection and response generation. Unlike existing datasets, it contains clean support knowledge, grounded at the utterance level and partitioned into multiple knowledge sources. We further propose a new challenge, dialogue knowledge plug-and-play, which aims to test an already trained dialogue model on using new support knowledge from previously unseen sources in a zero-shot fashion.

Sourced from various sensors and organized chronologically, Multivariate Time-Series (MTS) data involves crucial spatial-temporal dependencies, e.g., correlations among sensors. To capture these dependencies, Graph Neural Networks (GNNs) have emerged as powerful tools, yet their effectiveness is restricted by the quality of graph construction from MTS data. Typically, existing approaches construct graphs solely from MTS signals, which may introduce bias due to a small training dataset and may not accurately represent underlying dependencies. To address this challenge, we propose a novel framework named K-Link, leveraging Large Language Models (LLMs) to encode extensive general knowledge and thereby providing effective solutions to reduce the bias. Leveraging the knowledge embedded in LLMs, such as physical principles, we extract a \textit{Knowledge-Link graph}, capturing vast semantic knowledge of sensors and the linkage of the sensor-level knowledge. To harness the potential of the knowledge-link graph in enhancing the graph derived from MTS data, we propose a graph alignment module, facilitating the transfer of semantic knowledge within the knowledge-link graph into the MTS-derived graph. By doing so, we can improve the graph quality, ensuring effective representation learning with GNNs for MTS data. Extensive experiments demonstrate the efficacy of our approach for superior performance across various MTS-related downstream tasks.

Large Language Models (LLM) have shown encouraging progress in multimodal understanding and generation tasks. However, how to design a human-aligned and interpretable melody composition system is still under-explored. To solve this problem, we propose ByteComposer, an agent framework emulating a human's creative pipeline in four separate steps : "Conception Analysis - Draft Composition - Self-Evaluation and Modification - Aesthetic Selection". This framework seamlessly blends the interactive and knowledge-understanding features of LLMs with existing symbolic music generation models, thereby achieving a melody composition agent comparable to human creators. We conduct extensive experiments on GPT4 and several open-source large language models, which substantiate our framework's effectiveness. Furthermore, professional music composers were engaged in multi-dimensional evaluations, the final results demonstrated that across various facets of music composition, ByteComposer agent attains the level of a novice melody composer.

The superior performance of supervised classification methods in the information extraction (IE) area heavily relies on a large amount of gold standard data. Recent zero-shot classification methods converted the task to other NLP tasks (e.g., textual entailment) and used off-the-shelf models of these NLP tasks to directly perform inference on the test data without using a large amount of IE annotation data. A potentially valuable by-product of these methods is the large-scale silver standard data, i.e., pseudo-labeled data by the off-the-shelf models of other NLP tasks. However, there is no further investigation into the use of these data. In this paper, we propose a new framework, Clean-LaVe, which aims to utilize silver standard data to enhance the zero-shot performance. Clean-LaVe includes four phases: (1) Obtaining silver data; (2) Identifying relatively clean data from silver data; (3) Finetuning the off-the-shelf model using clean data; (4) Inference on the test data. The experimental results show that Clean-LaVe can outperform the baseline by 5% and 6% on TACRED and Wiki80 dataset in the zero-shot relation classification task, and by 3% ~7 % on Smile (Korean and Polish) in the zero-shot cross-lingual relation classification task, and by 8% on ACE05-E+ in the zero-shot event argument classification task.