Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, well before a model is even invoked. The same text translated into different languages can have drastically different tok-enization lengths, with differences up to 15 times in some cases. These disparities persist even for tokenizers that are intentionally trained for multilingual support.

Iran's supreme leader, Ayatollah Ali Khamenei, in his secret hideout these days, knows he is now a marked man. He will not be sitting on his veranda anytime soon. When discussing what the United States might do next to help the protesters in Iran, US President Trump has mentioned Qassem Soleimani and Abu Bakr al-Baghdadi. The former, Iran's all-important military strategist in the Middle East, was killed on 3 January 2020 in a drone strike just outside Baghdad's international airport on the president's order. The latter, who was the leader of IS, killed himself and two children by detonating a suicide vest on 27 October 2019 when US forces raided his hideout in northern Syria after the approval of the president.

Arabic is one of the world's most widely spoken languages with at least 400 million speakers, including 200 million native speakers and 200 million to 250 million non-native speakers. Modern Standard Arabic (MSA) serves as the formal language for government, legal matters and education, and it is widely used in international and religious contexts. Additionally, more than 25 dialects are spoken primarily across the Middle East and North Africa. The date was chosen to mark the day in 1973 on which the UN General Assembly adopted Arabic as one of its six official languages. In the following visual explainer, Al Jazeera lists some of the most common words in today's English language that originated from Arabic or passed through Arabic before reaching English.

Large language models have the potential to generate explanations for their own predictions in a variety of styles based on user instructions. Recent research has examined whether these self-explanations faithfully reflect the models' actual behavior and has found that they often lack faithfulness. However, the question of how to improve faithfulness remains underexplored. Moreover, because different explanation styles have superficially distinct characteristics, it is unclear whether improvements observed in one style also arise when using other styles. This study analyzes the effects of training for faithful self-explanations and the extent to which these effects generalize, using three classification tasks and three explanation styles. We construct one-word constrained explanations that are likely to be faithful using a feature attribution method, and use these pseudo-faithful self-explanations for continual learning on instruction-tuned models. Our experiments demonstrate that training can improve self-explanation faithfulness across all classification tasks and explanation styles, and that these improvements also show signs of generalization to the multi-word settings and to unseen tasks. Furthermore, we find consistent cross-style generalization among three styles, suggesting that training may contribute to a broader improvement in faithful self-explanation ability.

Boolean matrix factorization (BMF) approximates a given binary input matrix as the product of two smaller binary factors. Unlike binary matrix factorization based on standard arithmetic, BMF employs the Boolean OR and AND operations for the matrix product, which improves interpretability and reduces the approximation error. It is also used in role mining and computer vision. In this paper, we first propose algorithms for BMF that perform alternating optimization (AO) of the factor matrices, where each subproblem is solved via integer programming (IP). We then design different approaches to further enhance AO-based algorithms by selecting an optimal subset of rank-one factors from multiple runs. To address the scalability limits of IP-based methods, we introduce new greedy and local-search heuristics. We also construct a new C++ data structure for Boolean vectors and matrices that is significantly faster than existing ones and is of independent interest, allowing our heuristics to scale to large datasets. We illustrate the performance of all our proposed methods and compare them with the state of the art on various real datasets, both with and without missing data, including applications in topic modeling and imaging.

Neural Information Processing Systems http://nips.cc/

Large language models (LLMs) are increasingly used in finance and economics, where prompt-based attempts against look-ahead bias implicitly assume that models understand chronology. We test this fundamental question with a series of chronological ordering tasks with increasing complexities over facts the model already knows from pre-training. Our tasks cover (1) chronological ordering, (2) conditional sorting (filter, then order), and (3) anachronism detection. We evaluate GPT-4.1, Claude-3.7 Sonnet, with and without Extended Thinking (ET), and GPT-5 across multiple reasoning-effort settings. Across models, Exact match rate drops sharply as sequences lengthen even while rank correlations stay high as LLMs largely preserve local order but struggle to maintain a single globally consistent timeline. In conditional sorting, most failures stem from the filtering step rather than the ordering step, but GPT-5 and Claude-3.7 Sonnet with Extended Thinking outshine normal models significantly. Lastly, anachronism detection is found to be the easiest task for the LLMs but performance still declines with increasingly overlapping timelines or entities. Overall, our main contribution is showing that allocating explicit reasoning budget helps with chronological ordering with GPT-5 at medium/high reasoning effort achieving flawless ordering at all lengths and perfect conditional sorting (both self-filtered and given-subset), whereas low/minimal effort degrades with longer lists, mirroring earlier models. Our findings delineate limits of current LLMs on chronological tasks, providing insights into task complexity, and demonstrate scenarios in which reasoning helps. These patterns are important for the real-time application of LLMs in finance. We release all code and evaluation templates to support full reproducibility.

Current studies perceive LLMs' knowledge boundary

Accurate short-range prediction of extreme air temperature events remains a fundamental challenge in operational climate-risk management. We present Multi-Modal Weather State Transition Model with Anomaly-Driven Recurrent Attention Network Plus (MMWSTM-ADRAN+), a dual-stream deep learning architecture that couples a regime-aware dynamics model with an anomaly-focused attention mechanism to forecast daily maximum temperature and its extremes. The first stream, MMWSTM, combines bidirectional Long Short-Term Memory (BiLSTM) units with a learnable Markov state transition matrix to capture synoptic-scale weather regime changes. The second stream, ADRAN, integrates bidirectional Gated Recurrent Units (BiGRUs), multi-head self-attention, and a novel anomaly amplification layer to enhance sensitivity to low-probability signals. A lightweight attentive fusion gate adaptively determines the contribution of each stream to the final prediction. Model optimization employs a custom ExtremeWeatherLoss function that up-weights errors on the upper 5% and lower 5% of the temperature distribution, and a time-series data augmentation suite (jittering, scaling, time/magnitude warping) that effectively quadruples the training data

We present DialectalArabicMMLU, a new benchmark for evaluating the performance of large language models (LLMs) across Arabic dialects. While recently developed Arabic and multilingual benchmarks have advanced LLM evaluation for Modern Standard Arabic (MSA), dialectal varieties remain underrepresented despite their prevalence in everyday communication. DialectalArabicMMLU extends the MMLU-Redux framework through manual translation and adaptation of 3K multiple-choice question-answer pairs into five major dialects (Syrian, Egyptian, Emirati, Saudi, and Moroccan), yielding a total of 15K QA pairs across 32 academic and professional domains (22K QA pairs when also including English and MSA). The benchmark enables systematic assessment of LLM reasoning and comprehension beyond MSA, supporting both task-based and linguistic analysis. We evaluate 19 open-weight Arabic and multilingual LLMs (1B-13B parameters) and report substantial performance variation across dialects, revealing persistent gaps in dialectal generalization. DialectalArabicMMLU provides the first unified, human-curated resource for measuring dialectal understanding in Arabic, thus promoting more inclusive evaluation and future model development.