Millions of people in African countries face barriers to accessing healthcare due to language and literacy gaps. This research tackles this challenge by transforming complex medical documents -- in this case, prosthetic device user manuals -- into accessible formats for underserved populations. This case study in cross-cultural translation is particularly pertinent/relevant for communities that receive donated prosthetic devices but may not receive the accompanying user documentation. Or, if available online, may only be available in formats (e.g., language and readability) that are inaccessible to local populations (e.g., English-language, high resource settings/cultural context). The approach is demonstrated using the widely spoken Pidgin dialect, but our open-source framework has been designed to enable rapid and easy extension to other languages/dialects. This work presents an AI-powered framework designed to process and translate complex medical documents, e.g., user manuals for prosthetic devices, into marginalised languages. The system enables users -- such as healthcare workers or patients -- to upload English-language medical equipment manuals, pose questions in their native language, and receive accurate, localised answers in real time. Technically, the system integrates a Retrieval-Augmented Generation (RAG) pipeline for processing and semantic understanding of the uploaded manuals. It then employs advanced Natural Language Processing (NLP) models for generative question-answering and multilingual translation. Beyond simple translation, it ensures accessibility to device instructions, treatment protocols, and safety information, empowering patients and clinicians to make informed healthcare decisions.

Large Language Models (LLMs) have shown significant progress on various multilingual benchmarks and are increasingly used to generate and evaluate text in non-English languages. However, while they may produce fluent outputs, it remains unclear to what extent these models truly grasp the underlying linguistic complexity of those languages, particularly in morphology. To investigate this, we introduce IMPACT, a synthetically generated evaluation framework focused on inflectional morphology, which we publicly release, designed to evaluate LLM performance across five morphologically rich languages: Arabic, Russian, Finnish, Turkish, and Hebrew. IMPACT includes unit-test-style cases covering both shared and language-specific phenomena, from basic verb inflections (e.g., tense, number, gender) to unique features like Arabic's reverse gender agreement and vowel harmony in Finnish and Turkish. We assess eight multilingual LLMs that, despite strong English performance, struggle with other languages and uncommon morphological patterns, especially when judging ungrammatical examples. We also show that Chain of Thought and Thinking Models can degrade performance. Our work exposes gaps in LLMs' handling of linguistic complexity, pointing to clear room for improvement. To support further research, we publicly release the IMPACT framework.

Large language models (LLMs) have exhibited expert-level capabilities across various domains. However, their abilities to solve problems in Operations Research (OR) -- the analysis and optimization of mathematical models derived from real-world problems or their verbal descriptions -- remain underexplored. In this work, we take a first step toward evaluating LLMs' abilities to solve stochastic modeling problems, a core class of OR problems characterized by uncertainty and typically involving tools from probability, statistics, and stochastic processes. We manually procure a representative set of graduate-level homework and doctoral qualification-exam problems and test LLMs' abilities to solve them. We further leverage SimOpt, an open-source library of simulation-optimization problems and solvers, to investigate LLMs' abilities to make real-world decisions under uncertainty. Our results show that, though a nontrivial amount of work is still needed to reliably automate the stochastic modeling pipeline in reality, state-of-the-art LLMs demonstrate proficiency on par with human experts in both classroom and practical settings. These findings highlight the potential of building AI agents that assist OR researchers and amplify the real-world impact of OR through automation.

The chain of thought is fundamental in Transformers, which is to perform step-by-step reasoning. Besides what intermediate steps work, the order of these steps critically affects the difficulty of the reasoning. This study addresses a novel task of unraveling chain of thought - reordering decoder input tokens to a learning-friendly sequence for Transformers to learn arithmetic tasks. The proposed pipeline first trains a Transformer on a mixture of target sequences arranged in different orders and then identifies benign orders as those with fast loss drops in the early stage. As the search space grows factorially with sequence length, we propose a two-stage hierarchical approach for inter- and intra-block reordering. Experiments on four order-sensitive arithmetic tasks show that our method identifies a learning-friendly order out of a few billion candidates. Notably, on the multiplication task, it recovered the reverse-digit order reported in prior studies.

We study the capabilities of generative autoregressive transformer models trained on large amounts of symbolic solo-piano transcriptions. After first pretraining on approximately 60,000 hours of music, we use a comparatively smaller, high-quality subset, to finetune models to produce musical continuations, perform symbolic classification tasks, and produce general-purpose contrastive MIDI embeddings by adapting the SimCLR framework to symbolic music. When evaluating piano continuation coherence, our generative model outperforms leading symbolic generation techniques and remains competitive with proprietary audio generation models. On MIR classification benchmarks, frozen representations from our contrastive model achieve state-of-the-art results in linear probe experiments, while direct finetuning demonstrates the generalizability of pretrained representations, often requiring only a few hundred labeled examples to specialize to downstream tasks.

We conduct a systematic audit of three widely used reasoning benchmarks, SocialIQa, FauxPas-EAI, and ToMi, and uncover pervasive flaws in both benchmark items and evaluation methodology. Using five LLMs (GPT-{3, 3.5, 4, o1}, and LLaMA 3.1) as diagnostic tools, we identify structural, semantic, and pragmatic issues in benchmark design (e.g., duplicated items, ambiguous wording, and implausible answers), as well as scoring procedures that prioritize output form over reasoning process. Through systematic human annotation and re-evaluation on cleaned benchmark subsets, we find that model scores often improve not due to due to erratic surface wording variations and not to improved reasoning. Infact, further analyses show that model performance is highly sensitive to minor input variations such as context availability and phrasing, revealing that high scores may reflect alignment with format-specific cues rather than consistent inference based on the input. These findings challenge the validity of current benchmark-based claims about reasoning in LLMs, and highlight the need for evaluation protocols that assess reasoning as a process of drawing inference from available information, rather than as static output selection. We release audited data and evaluation tools to support more interpretable and diagnostic assessments of model reasoning.

Large Reasoning Models (LRMs) excel at solving complex problems but face an overthinking dilemma. When handling simple tasks, they often produce verbose responses overloaded with thinking tokens (e.g., wait, however). These tokens trigger unnecessary high-level reasoning behaviors like reflection and backtracking, reducing efficiency. In this work, our pilot study reveals that these thinking-token-induced behaviors are not essential for effective problem-solving and may even hinder correct reasoning within constrained token budgets. We identify this phenomenon as the thinking trap. To mitigate this issue, we propose Dual Policy Preference Optimization (DuP-PO), a novel algorithm featuring: (1) A rollout sampling strategy that guarantees balanced exposure to responses with and without thinking tokens; (2) A fine-grained advantage control technique to dynamically regulate the prediction of target tokens; (3) A policy shaping method ensuring stable gradient contributions from thinking tokens. Experimental results on five popular math reasoning benchmarks show that DuP-PO performs well on the popular LRM, which significantly improves their token efficiency during reasoning, while achieving superior performance of the base model.

External tool integration through function-calling is essential for practical language model applications, yet most multilingual models lack reliable tool-use capabilities in non-English languages. Even state-of-the-art multilingual models struggle with determining when to use tools and generating the structured outputs required for function calls, often exhibiting language confusion when prompted in lower-resource languages. This work presents a methodology for adapting existing language models to enable robust tool use in any target language, using Bulgarian as a case study. The approach involves continued training of the BgGPT model series (2.6B, 9B, 27B parameters) on a novel bilingual dataset of 10,035 function-calling examples designed to support standardized protocols like MCP (Model Context Protocol). The research introduces TUCAN (Tool-Using Capable Assistant Navigator), which achieves up to 28.75% improvement in function-calling accuracy over base models while preserving core language understanding, as verified on established Bulgarian benchmarks. Beyond accuracy gains, TUCAN models demonstrate production-ready response formatting with clean, parsable function calls, contrasting with the verbose and inconsistent outputs of base models. The models, evaluation framework, and dataset are released to enable replication for other languages. This work demonstrates a practical approach for extending tool-augmented capabilities beyond English-centric systems.

Brain and Mind Institute, The Chinese University of Hong Kong, Hong Kong, Hong Kong SAR, China. Correspondence should be addressed to Zhenguang G. Cai, Department of Linguistics and Modern Languages, Leung Kau Kui Building, The Chinese University of Hong Kong, Shatin, Hong Kong SAR; zhenguangcai@cuhk.edu.hk. Abstract: Large l anguage m odels have achieved impressive progress in multilingual translation, yet they continue to face challenges with certain language pairs --particularly those with limited training data or significant linguistic divergence from English. This study systematically investigates how training data, language proximity, and language family affect information loss in multilingual translation . We evaluate two large language model s, GPT - 4 and Llama 2, by performing round-trip translation s . Translation quality was assessed using BLEU scores and BERT similarity metrics. Our results reveal a robust interaction between training data size and language distance: while abundant training data can mitigate the effects of linguistic divergence, languages structurally closer to English consistently yield higher translation quality in low - resource conditions. Among various distance metrics, orthographic, phylogenetic, syntactic, and geographical distances emerge as strong predictors of translation performance. L anguage family also exert s an independent influence. These findings contribute to a deeper understanding of the linguistic constraints shaping multilingual translation in large language models, emphasizing that translation quality is shaped not only by data volume but also by structural and typological relationships between languages. 1 INTRODUCTION Large Language Models (LLMs) demonstrated advanced multilingual capabilities.

The development of generalist robot manipulation policies has seen significant progress, driven by large-scale demonstration data across diverse environments. However, the high cost and inefficiency of collecting real-world demonstrations hinder the scalability of data acquisition. While existing simulation platforms enable controlled environments for robotic learning, the challenge of bridging the sim-to-real gap remains. To address these challenges, we propose RoboPearls, an editable video simulation framework for robotic manipulation. Built on 3D Gaussian Splatting (3DGS), RoboPearls enables the construction of photo-realistic, view-consistent simulations from demonstration videos, and supports a wide range of simulation operators, including various object manipulations, powered by advanced modules like Incremental Semantic Distillation (ISD) and 3D regularized NNFM Loss (3D-NNFM). Moreover, by incorporating large language models (LLMs), RoboPearls automates the simulation production process in a user-friendly manner through flexible command interpretation and execution. Furthermore, RoboPearls employs a vision-language model (VLM) to analyze robotic learning issues to close the simulation loop for performance enhancement. To demonstrate the effectiveness of RoboPearls, we conduct extensive experiments on multiple datasets and scenes, including RLBench, COLOSSEUM, Ego4D, Open X-Embodiment, and a real-world robot, which demonstrate our satisfactory simulation performance.