Event Extraction (EE) is an essential information extraction task that aims to extract event-related information from unstructured texts. The paradigm of this task has shifted from conventional classification-based methods to more contemporary question-answering-based (QA-based) approaches. However, in QA-based EE, the quality of the questions dramatically affects the extraction accuracy, and how to generate high-quality questions for QA-based EE remains a challenge. In this work, to tackle this challenge, we suggest four criteria to evaluate the quality of a question and propose a reinforcement learning method, RLQG, for QA-based EE that can generate generalizable, high-quality, and context-dependent questions and provides clear guidance to QA models. The extensive experiments conducted on ACE and RAMS datasets have strongly validated our approach's effectiveness, which also demonstrates its robustness in scenarios with limited training data. The corresponding code of RLQG is released for further research.

Advancements in large language models (LLMs) are revolutionizing interactive game design, enabling dynamic plotlines and interactions between players and non-player characters (NPCs). However, LLMs may exhibit flaws such as hallucinations, forgetfulness, or misinterpretations of prompts, causing logical inconsistencies and unexpected deviations from intended designs. Automated techniques for detecting such game bugs are still lacking. To address this, we propose a systematic LLM-based method for automatically identifying such bugs from player game logs, eliminating the need for collecting additional data such as post-play surveys. Applied to a text-based game DejaBoom!, our approach effectively identifies bugs inherent in LLM-powered interactive games, surpassing unstructured LLM-powered bug-catching methods and filling the gap in automated detection of logical and design flaws.

We empirically demonstrate that a transformer pre-trained on country-scale unlabeled human mobility data learns embeddings capable, through fine-tuning, of developing a deep understanding of the target geography and its corresponding mobility patterns. Utilizing an adaptation framework, we evaluate the performance of our pre-trained embeddings in encapsulating a broad spectrum of concepts directly and indirectly related to human mobility. This includes basic notions, such as geographic location and distance, and extends to more complex constructs, such as administrative divisions and land cover. Our extensive empirical analysis reveals a substantial performance boost gained from pre-training, reaching up to 38% in tasks such as tree-cover regression. We attribute this result to the ability of the pre-training to uncover meaningful patterns hidden in the raw data, beneficial for modeling relevant Figure 1: A transformer pre-trained from scratch on countryscale high-level concepts. The pre-trained embeddings emerge as robust unlabeled human mobility data is adapted to model a representations of regions and trajectories, potentially valuable for variety of high-level concepts manifesting at different levels a wide range of downstream applications.

The advancements in embodied AI are increasingly enabling robots to tackle complex real-world tasks, such as household manipulation. However, the deployment of robots in these environments remains constrained by the lack of comprehensive bimanual-mobile robot manipulation data that can be learned. Existing datasets predominantly focus on single-arm manipulation tasks, while the few dual-arm datasets available often lack mobility features, task diversity, comprehensive sensor data, and robust evaluation metrics; they fail to capture the intricate and dynamic nature of household manipulation tasks that bimanual-mobile robots are expected to perform. To overcome these limitations, we propose BRMData, a Bimanual-mobile Robot Manipulation Dataset specifically designed for household applications. BR-MData encompasses 10 diverse household tasks, including single-arm and dual-arm tasks, as well as both tabletop and mobile manipulations, utilizing multi-view and depth-sensing data information. Moreover, BRMData features tasks of increasing difficulty, ranging from single-object to multi-object grasping, non-interactive to human-robot interactive scenarios, and rigid-object to flexible-object manipulation, closely simulating real-world household applications. Additionally, we introduce a novel Manipulation Efficiency Score (MES) metric to evaluate both the precision and efficiency of robot manipulation methods in household tasks. We thoroughly evaluate and analyze the performance of advanced robot manipulation learning methods using our BRMData, aiming to drive the development of bimanual-mobile robot manipulation technologies. The dataset is now open-sourced and available at https://embodiedrobot.github.io/.

Large language models (LLMs) allow us to generate high-quality human-like text. One interesting task in natural language processing (NLP) is named entity recognition (NER), which seeks to detect mentions of relevant information in documents. This paper presents llmNER, a Python library for implementing zero-shot and few-shot NER with LLMs; by providing an easy-to-use interface, llmNER can compose prompts, query the model, and parse the completion returned by the LLM. Also, the library enables the user to perform prompt engineering efficiently by providing a simple interface to test multiple variables. We validated our software on two NER tasks to show the library's flexibility. llmNER aims to push the boundaries of in-context learning research by removing the barrier of the prompting and parsing steps.

In recent years there has been a tremendous surge in the general capabilities of AI systems, mainly fuelled by training foundation models on internetscale data. Nevertheless, the creation of openended, ever self-improving AI remains elusive. In this position paper, we argue that the ingredients are now in place to achieve openendedness in AI systems with respect to a human observer. Furthermore, we claim that such open-endedness is an essential property of any artificial superhuman intelligence (ASI). We begin by providing a concrete formal definition of open-endedness through the lens of novelty and learnability. We then illustrate a path towards ASI via open-ended systems built on top of foundation models, capable of making novel, humanrelevant discoveries. We conclude by examining the safety implications of generally-capable openended AI. We expect that open-ended foundation models will prove to be an increasingly fertile and safety-critical area of research in the near future.

Understanding memorisation in language models has practical and societal implications, e.g., studying models' training dynamics or preventing copyright infringements. Prior work defines memorisation as the causal effect of training with an instance on the model's ability to predict that instance. This definition relies on a counterfactual: the ability to observe what would have happened had the model not seen that instance. Existing methods struggle to provide computationally efficient and accurate estimates of this counterfactual. Further, they often estimate memorisation for a model architecture rather than for a specific model instance. This paper fills an important gap in the literature, proposing a new, principled, and efficient method to estimate memorisation based on the difference-in-differences design from econometrics. Using this method, we characterise a model's memorisation profile--its memorisation trends across training--by only observing its behaviour on a small set of instances throughout training. In experiments with the Pythia model suite, we find that memorisation (i) is stronger and more persistent in larger models, (ii) is determined by data order and learning rate, and (iii) has stable trends across model sizes, thus making memorisation in larger models predictable from smaller ones.

The advancement of large language models (LLMs) relies on evaluation using public benchmarks, but data contamination can lead to overestimated performance. Previous researches focus on detecting contamination by determining whether the model has seen the exact same data during training. In this work, we argue that even training on data similar to benchmark data inflates performance on in-distribution tasks without improving overall capacity, which we called In-distribution contamination. To effectively detect in-distribution contamination, we propose DICE, a novel method that leverages the internal states of LLMs to locate-then-detect the contamination. DICE first identifies the most sensitive layer to contamination, then trains a classifier based on the internal states of that layer. Experiments reveal DICE's high accuracy in detecting in-distribution contamination across various LLMs and math reasoning datasets. We also show the generalization capability of the trained DICE detector, which is able to detect contamination across multiple benchmarks with similar distributions. Additionally, we find that the DICE detection scores are positively correlated with the performance of ten LLMs fine-tuned by either us or other organizations on four math reasoning datasets (with $R^2$ values between 0.6 and 0.75). This indicates that the in-distribution contamination problem potentially lead to an overestimation of the true capabilities of many existing models. The code and data are available at https://github.com/THU-KEG/DICE.

Large language models (LLMs) can suffer from hallucinations when generating text. These hallucinations impede various applications in society and industry by making LLMs untrustworthy. Current LLMs generate text in an autoregressive fashion by predicting and appending text tokens. When an LLM is uncertain about the semantic meaning of the next tokens to generate, it is likely to start hallucinating. Thus, it has been suggested that hallucinations stem from predictive uncertainty. We introduce Semantically Diverse Language Generation (SDLG) to quantify predictive uncertainty in LLMs. SDLG steers the LLM to generate semantically diverse yet likely alternatives for an initially generated text. This approach provides a precise measure of aleatoric semantic uncertainty, detecting whether the initial text is likely to be hallucinated. Experiments on question-answering tasks demonstrate that SDLG consistently outperforms existing methods while being the most computationally efficient, setting a new standard for uncertainty estimation in LLMs.

In this work, we systematically study music generation conditioned solely on the video. First, we present a large-scale dataset comprising 190K video-music pairs, including various genres such as movie trailers, advertisements, and documentaries. Furthermore, we propose VidMuse, a simple framework for generating music aligned with video inputs. VidMuse stands out by producing high-fidelity music that is both acoustically and semantically aligned with the video. By incorporating local and global visual cues, VidMuse enables the creation of musically coherent audio tracks that consistently match the video content through Long-Short-Term modeling. Through extensive experiments, VidMuse outperforms existing models in terms of audio quality, diversity, and audio-visual alignment. The code and datasets will be available at https://github.com/ZeyueT/VidMuse/.