The documentation of the Neo-Aramaic dialects before their extinction has been described as the most urgent task in all of Semitology today. The death of this language will be an unfathomable loss to the descendents of the indigenous speakers of Aramaic, now predominantly diasporic after forced displacement due to violence. This paper develops an ASR model to expedite the documentation of this endangered language and generalizes the strategy in a new framework we call NoLoR.

Language model post-training is applied to refine behaviors and unlock new skills across a wide range of recent language models, but open recipes for applying these techniques lag behind proprietary ones. The underlying training data and recipes for post-training are simultaneously the most important pieces of the puzzle and the portion with the least transparency. To bridge this gap, we introduce Tulu 3, a family of fully-open state-of-the-art post-trained models, alongside its data, code, and training recipes, serving as a comprehensive guide for modern post-training techniques. Tulu 3, which builds on Llama 3.1 base models, achieves results surpassing the instruct versions of Llama 3.1, Qwen 2.5, Mistral, and even closed models such as GPT-4o-mini and Claude 3.5-Haiku. The training algorithms for our models include supervised finetuning (SFT), Direct Preference Optimization (DPO), and a novel method we call Reinforcement Learning with Verifiable Rewards (RLVR). With Tulu 3, we introduce a multi-task evaluation scheme for post-training recipes with development and unseen evaluations, standard benchmark implementations, and substantial decontamination of existing open datasets on said benchmarks. We conclude with analysis and discussion of training methods that did not reliably improve performance. In addition to the Tulu 3 model weights and demo, we release the complete recipe -- including datasets for diverse core skills, a robust toolkit for data curation and evaluation, the training code and infrastructure, and, most importantly, a detailed report for reproducing and further adapting the Tulu 3 approach to more domains.

Temporal action segmentation and long-term action anticipation are two popular vision tasks for the temporal analysis of actions in videos. Despite apparent relevance and potential complementarity, these two problems have been investigated as separate and distinct tasks. In this work, we tackle these two problems, action segmentation and action anticipation, jointly using a unified diffusion model dubbed ActFusion. The key idea to unification is to train the model to effectively handle both visible and invisible parts of the sequence in an integrated manner; the visible part is for temporal segmentation, and the invisible part is for future anticipation. To this end, we introduce a new anticipative masking strategy during training in which a late part of the video frames is masked as invisible, and learnable tokens replace these frames to learn to predict the invisible future. Experimental results demonstrate the bi-directional benefits between action segmentation and anticipation. ActFusion achieves the state-of-the-art performance across the standard benchmarks of 50 Salads, Breakfast, and GTEA, outperforming task-specific models in both of the two tasks with a single unified model through joint learning.

This paper introduces the counter-intuitive generalization results of overfitting pre-trained large language models (LLMs) on very small datasets. In the setting of open-ended text generation, it is well-documented that LLMs tend to generate repetitive and dull sequences, a phenomenon that is especially apparent when generating using greedy decoding. This issue persists even with state-of-the-art LLMs containing billions of parameters, trained via next-token prediction on large datasets. We find that by further fine-tuning these models to achieve a near-zero training loss on a small set of samples - a process we refer to as hyperfitting - the long-sequence generative capabilities are greatly enhanced. Greedy decoding with these Hyperfitted models even outperform Top-P sampling over long-sequences, both in terms of diversity and human preferences. This phenomenon extends to LLMs of various sizes, different domains, and even autoregressive image generation. We further find this phenomena to be distinctly different from that of Grokking and double descent. Surprisingly, our experiments indicate that hyperfitted models rarely fall into repeating sequences they were trained on, and even explicitly blocking these sequences results in high-quality output. All hyperfitted models produce extremely low-entropy predictions, often allocating nearly all probability to a single token. Despite the recent rapid advancements in artificial intelligence spearheaded by Transformer-based large language models (LLMs) and their emergent phenomena (Wei et al., 2022b; Bubeck et al., 2023), models trained on next-token pre-training objectives often degenerate when producing longer texts. This is particularly true for greedy decoding, and has resulted in mitigation strategies such as repetition penalties (Keskar et al., 2019) and nucleus sampling (Holtzman et al., 2020). However, when removing these heuristics and simply picking the top-1 candidate at each time-step, LLMs display strong tendencies to repeat themselves at the token, phrase, and sentence level (Holtzman et al., 2020), as is exemplified in Figure 1. This is a recurrent phenomenon for which there are many proposed hypotheses but, to the best of our knowledge, no definitive explanation exists. Color indicating how repetitive the generated text is. Although these models achieve significantly worse validation loss, they produce texts that align markedly better with human preferences and automatic diversity metrics. Indeed, we find that hyperfitting state-of-the-art LLMs yields capabilities that outperform models with 10x the number of parameters.

As the strength of Large Language Models (LLMs) has grown over recent years, so too has interest in their use as the underlying models for autonomous agents. Although LLMs demonstrate emergent abilities and broad expertise across natural language domains, their inherent unpredictability makes the implementation of LLM agents challenging, resulting in a gap between related research and the real-world implementation of such systems. To bridge this gap, this paper frames actionable insights and considerations from the research community in the context of established application paradigms to enable the construction and facilitate the informed deployment of robust LLM agents. Namely, we position relevant research findings into four broad categories--Planning, Memory, Tools, and Control Flow--based on common practices in application-focused literature and highlight practical considerations to make when designing agentic LLMs for real-world applications, such as handling stochasticity and managing resources efficiently. While we do not conduct empirical evaluations, we do provide the necessary background for discussing critical aspects of agentic LLM designs, both in academia and industry.

With increasing numbers of vulnerabilities exposed on the internet, autonomous penetration testing (pentesting) has emerged as an emerging research area, while reinforcement learning (RL) is a natural fit for studying autonomous pentesting. Previous research in RL-based autonomous pentesting mainly focused on enhancing agents' learning efficacy within abstract simulated training environments. They overlooked the applicability and generalization requirements of deploying agents' policies in real-world environments that differ substantially from their training settings. In contrast, for the first time, we shift focus to the pentesting agents' ability to generalize across unseen real environments. For this purpose, we propose a Generalizable Autonomous Pentesting framework (namely GAP) for training agents capable of drawing inferences from one to another -- a key requirement for the broad application of autonomous pentesting and a hallmark of human intelligence. GAP introduces a Real-to-Sim-to-Real pipeline with two key methods: domain randomization and meta-RL learning. Specifically, we are among the first to apply domain randomization in autonomous pentesting and propose a large language model-powered domain randomization method for synthetic environment generation. We further apply meta-RL to improve the agents' generalization ability in unseen environments by leveraging the synthetic environments. The combination of these two methods can effectively bridge the generalization gap and improve policy adaptation performance. Experiments are conducted on various vulnerable virtual machines, with results showing that GAP can (a) enable policy learning in unknown real environments, (b) achieve zero-shot policy transfer in similar environments, and (c) realize rapid policy adaptation in dissimilar environments.

Numerous politicians use social media platforms, particularly X, to engage with their constituents. This interaction allows constituents to pose questions and offer feedback but also exposes politicians to a barrage of hostile responses, especially given the anonymity afforded by social media. They are typically targeted in relation to their governmental role, but the comments also tend to attack their personal identity. This can discredit politicians and reduce public trust in the government. It can also incite anger and disrespect, leading to offline harm and violence. While numerous models exist for detecting hostility in general, they lack the specificity required for political contexts. Furthermore, addressing hostility towards politicians demands tailored approaches due to the distinct language and issues inherent to each country (e.g., Brexit for the UK). To bridge this gap, we construct a dataset of 3,320 English tweets spanning a two-year period manually annotated for hostility towards UK MPs. Our dataset also captures the targeted identity characteristics (race, gender, religion, none) in hostile tweets. We perform linguistic and topical analyses to delve into the unique content of the UK political data. Finally, we evaluate the performance of pre-trained language models and large language models on binary hostility detection and multi-class targeted identity type classification tasks. Our study offers valuable data and insights for future research on the prevalence and nature of politics-related hostility specific to the UK.

Game-theoretic scenarios have become pivotal in evaluating the social intelligence of Large Language Model (LLM)-based social agents. While numerous studies have explored these agents in such settings, there is a lack of a comprehensive survey summarizing the current progress. To address this gap, we systematically review existing research on LLM-based social agents within game-theoretic scenarios. Our survey organizes the findings into three core components: Game Framework, Social Agent, and Evaluation Protocol. The game framework encompasses diverse game scenarios, ranging from choice-focusing to communication-focusing games. The social agent part explores agents' preferences, beliefs, and reasoning abilities. The evaluation protocol covers both game-agnostic and game-specific metrics for assessing agent performance. By reflecting on the current research and identifying future research directions, this survey provides insights to advance the development and evaluation of social agents in game-theoretic scenarios.

Estimating individual treatment effects (ITE) from observational data is a critical task across various domains. However, many existing works on ITE estimation overlook the influence of hidden confounders, which remain unobserved at the individual unit level. To address this limitation, researchers have utilized graph neural networks to aggregate neighbors' features to capture the hidden confounders and mitigate confounding bias by minimizing the discrepancy of confounder representations between the treated and control groups. Despite the success of these approaches, practical scenarios often treat all features as confounders and involve substantial differences in feature distributions between the treated and control groups. Confusing the adjustment and confounder and enforcing strict balance on the confounder representations could potentially undermine the effectiveness of outcome prediction. To mitigate this issue, we propose a novel framework called the \textit{Graph Disentangle Causal model} (GDC) to conduct ITE estimation in the network setting. GDC utilizes a causal disentangle module to separate unit features into adjustment and confounder representations. Then we design a graph aggregation module consisting of three distinct graph aggregators to obtain adjustment, confounder, and counterfactual confounder representations. Finally, a causal constraint module is employed to enforce the disentangled representations as true causal factors. The effectiveness of our proposed method is demonstrated by conducting comprehensive experiments on two networked datasets.

The rapid advancement of deep learning has resulted in substantial advancements in AI-driven applications; however, the "black box" characteristic of these models frequently constrains their interpretability, transparency, and reliability. Explainable artificial intelligence (XAI) seeks to elucidate AI decision-making processes, guaranteeing that explanations faithfully represent the model's rationale and correspond with human comprehension. Despite comprehensive research in XAI, a significant gap persists in standardized procedures for assessing the efficacy and transparency of XAI techniques across many real-world applications. This study presents a unified XAI evaluation framework incorporating extensive quantitative and qualitative criteria to systematically evaluate the correctness, interpretability, robustness, fairness, and completeness of explanations generated by AI models. The framework prioritizes user-centric and domain-specific adaptations, hence improving the usability and reliability of AI models in essential domains. To address deficiencies in existing evaluation processes, we suggest defined benchmarks and a systematic evaluation pipeline that includes data loading, explanation development, and thorough method assessment. The suggested framework's relevance and variety are evidenced by case studies in healthcare, finance, agriculture, and autonomous systems. These provide a solid basis for the equitable and dependable assessment of XAI methodologies. This paradigm enhances XAI research by offering a systematic, flexible, and pragmatic method to guarantee transparency and accountability in AI systems across many real-world contexts.