The widespread adoption of large language models (LLMs) has created an urgent need for robust tools to detect LLM-generated text, especially in light of \textit{paraphrasing} techniques that often evade existing detection methods. To address this challenge, we present a novel semantic-enhanced framework for detecting LLM-generated text (SEFD) that leverages a retrieval-based mechanism to fully utilize text semantics. Our framework improves upon existing detection methods by systematically integrating retrieval-based techniques with traditional detectors, employing a carefully curated retrieval mechanism that strikes a balance between comprehensive coverage and computational efficiency. We showcase the effectiveness of our approach in sequential text scenarios common in real-world applications, such as online forums and Q\&A platforms. Through comprehensive experiments across various LLM-generated texts and detection methods, we demonstrate that our framework substantially enhances detection accuracy in paraphrasing scenarios while maintaining robustness for standard LLM-generated content.

Artificial Intelligence (AI) has demonstrated significant capabilities in various fields, and in areas such as human-computer interaction (HCI), embodied intelligence, and the design and animation of virtual digital humans, both practitioners and users are increasingly concerned with AI's ability to understand and express emotion. Consequently, the question of whether AI can accurately interpret human emotions remains a critical challenge. To date, two primary classes of AI models have been involved in human emotion analysis: generative models and Multimodal Large Language Models (MLLMs). To assess the emotional capabilities of these two classes of models, this study introduces MEMO-Bench, a comprehensive benchmark consisting of 7,145 portraits, each depicting one of six different emotions, generated by 12 Text-to-Image (T2I) models. Unlike previous works, MEMO-Bench provides a framework for evaluating both T2I models and MLLMs in the context of sentiment analysis. Additionally, a progressive evaluation approach is employed, moving from coarse-grained to fine-grained metrics, to offer a more detailed and comprehensive assessment of the sentiment analysis capabilities of MLLMs. The experimental results demonstrate that existing T2I models are more effective at generating positive emotions than negative ones. Meanwhile, although MLLMs show a certain degree of effectiveness in distinguishing and recognizing human emotions, they fall short of human-level accuracy, particularly in fine-grained emotion analysis. The MEMO-Bench will be made publicly available to support further research in this area.

Watermarking language models is essential for distinguishing between human and machine-generated text and thus maintaining the integrity and trustworthiness of digital communication. We present a novel green/red list watermarking approach that partitions the token set into ``green'' and ``red'' lists, subtly increasing the generation probability for green tokens. To correct token distribution bias, our method employs maximal coupling, using a uniform coin flip to decide whether to apply bias correction, with the result embedded as a pseudorandom watermark signal. Theoretical analysis confirms this approach's unbiased nature and robust detection capabilities. Experimental results show that it outperforms prior techniques by preserving text quality while maintaining high detectability, and it demonstrates resilience to targeted modifications aimed at improving text quality. This research provides a promising watermarking solution for language models, balancing effective detection with minimal impact on text quality.

Language models trained on diverse datasets unlock generalization by in-context learning. Reinforcement Learning (RL) policies can achieve a similar effect by meta-learning within the memory of a sequence model. However, meta-RL research primarily focuses on adapting to minor variations of a single task. It is difficult to scale towards more general behavior without confronting challenges in multi-task optimization, and few solutions are compatible with meta-RL's goal of learning from large training sets of unlabeled tasks. To address this challenge, we revisit the idea that multi-task RL is bottlenecked by imbalanced training losses created by uneven return scales across different tasks. We build upon recent advancements in Transformer-based (in-context) meta-RL and evaluate a simple yet scalable solution where both an agent's actor and critic objectives are converted to classification terms that decouple optimization from the current scale of returns. Large-scale comparisons in Meta-World ML45, Multi-Game Procgen, Multi-Task POPGym, Multi-Game Atari, and BabyAI find that this design unlocks significant progress in online multi-task adaptation and memory problems without explicit task labels.

Sparse Autoencoders for transformer-based language models are typically defined independently per layer. In this work we analyze statistical relationships between features in adjacent layers to understand how features evolve through a forward pass. We provide a graph visualization interface for features and their most similar next-layer neighbors, and build communities of related features across layers. We find that a considerable amount of features are passed through from a previous layer, some features can be expressed as quasi-boolean combinations of previous features, and some features become more specialized in later layers.

Textual description of a physical location, commonly known as an address, plays an important role in location-based services(LBS) such as on-demand delivery and navigation. However, the prevalence of abnormal addresses, those containing inaccuracies that fail to pinpoint a location, have led to significant costs. Address rewriting has emerged as a solution to rectify these abnormal addresses. Despite the critical need, existing address rewriting methods are limited, typically tailored to correct specific error types, or frequently require retraining to process new address data effectively. In this study, we introduce AddrLLM, an innovative framework for address rewriting that is built upon a retrieval augmented large language model. AddrLLM overcomes aforementioned limitations through a meticulously designed Supervised Fine-Tuning module, an Address-centric Retrieval Augmented Generation module and a Bias-free Objective Alignment module. To the best of our knowledge, this study pioneers the application of LLM-based address rewriting approach to solve the issue of abnormal addresses. Through comprehensive offline testing with real-world data on a national scale and subsequent online deployment, AddrLLM has demonstrated superior performance in integration with existing logistics system. It has significantly decreased the rate of parcel re-routing by approximately 43\%, underscoring its exceptional efficacy in real-world applications.

Reasoning about time and temporal relations is an integral aspect of human cognition, essential for perceiving the world and navigating our experiences. Though large language models (LLMs) have demonstrated impressive performance in many reasoning tasks, temporal reasoning remains challenging due to its intrinsic complexity. In this work, we first study an essential task of temporal reasoning -- temporal graph generation, to unveil LLMs' inherent, global reasoning capabilities. We show that this task presents great challenges even for the most powerful LLMs, such as GPT-3.5/4. We also notice a significant performance gap by small models (<10B) that lag behind LLMs by 50%. Next, we study how to close this gap with a budget constraint, e.g., not using model finetuning. We propose a new prompting technique tailored for temporal reasoning, Narrative-of-Thought (NoT), that first converts the events set to a Python class, then prompts a small model to generate a temporally grounded narrative, guiding the final generation of a temporal graph. Extensive experiments showcase the efficacy of NoT in improving various metrics. Notably, NoT attains the highest F1 on the Schema-11 evaluation set, while securing an overall F1 on par with GPT-3.5. NoT also achieves the best structural similarity across the board, even compared with GPT-3.5/4. Our code is available at https://github.com/launchnlp/NoT.

Abstract: Much linguistic research relies on annotated datasets of features extracted from text corpora, but the rapid quantitative growth of these corpora has created practical difficulties for linguists to manually annotate large data samples. In this paper, we present a replicable, supervised method that leverages large language models for assisting the linguist in grammatical annotation through prompt engineering, training, and evaluation. We introduce a methodological pipeline applied to the case study of formal variation in the English evaluative verb construction'consider X (as) (to be) Y', based on the large language model Claude 3.5 Sonnet and corpus data from Davies' NOW and EnTenTen21 (SketchEngine). Overall, we reach a model accuracy of over 90% on our held-out test samples with only a small amount of training data, validating the method for the annotation of very large quantities of tokens of the construction in the future. We discuss the generalisability of our results for a wider range of case studies of grammatical constructions and grammatical variation and change, underlining the value of AI copilots as tools for future linguistic research. Keywords: corpus linguistics; grammar; artificial intelligence; large language models; annotation 1. Introduction Corpus linguistic research typically manoeuvres vast quantities of data, which appear to have only kept growing since the 1990s and the early 2000s in the context of the'quantitative turn' undergone by the field (Kortmann 2021).

High annotation costs from hiring or crowdsourcing complicate the creation of large, high-quality datasets needed for training reliable text classifiers. Recent research suggests using Large Language Models (LLMs) to automate the annotation process, reducing these costs while maintaining data quality. LLMs have shown promising results in annotating downstream tasks like hate speech detection and political framing. Building on the success in these areas, this study investigates whether LLMs are viable for annotating the complex task of media bias detection and whether a downstream media bias classifier can be trained on such data. We create annolexical, the first large-scale dataset for media bias classification with over 48000 synthetically annotated examples. Our classifier, fine-tuned on this dataset, surpasses all of the annotator LLMs by 5-9 percent in Matthews Correlation Coefficient (MCC) and performs close to or outperforms the model trained on human-labeled data when evaluated on two media bias benchmark datasets (BABE and BASIL). This study demonstrates how our approach significantly reduces the cost of dataset creation in the media bias domain and, by extension, the development of classifiers, while our subsequent behavioral stress-testing reveals some of its current limitations and trade-offs.

Candice Lim is joined by Dazed senior writer Laura Pitcher to discuss her piece, "Meet the people using ChatGPT as their therapist." Since ChatGPT's public release in late 2022, there has been a growing reliance on the artificially intelligent chatbot in people's everyday lives. TikTok users are talking about the way they use ChatGPT as their therapist, their best friend, their life organizer, and more. But is reliant, daily use of an AI service worth possible consequences such as climate change, loneliness, and data privacy concerns? On today's episode, ICYMI dives into the unexpected ways people have been using ChatGPT and whether we can predict its role in the near future.