Chain-of-thought (CoT) reasoning has been widely applied in the mathematical reasoning of Large Language Models (LLMs). Recently, the introduction of derivative process supervision on CoT trajectories has sparked discussions on enhancing scaling capabilities during test time, thereby boosting the potential of these models. However, in multimodal mathematical reasoning, the scarcity of high-quality CoT training data has hindered existing models from achieving high-precision CoT reasoning and has limited the realization of reasoning potential during test time. In this work, we propose a three-module synthesis strategy that integrates CoT distillation, trajectory-format rewriting, and format unification. It results in a high-quality CoT reasoning instruction fine-tuning dataset in multimodal mathematics, MMathCoT-1M. We comprehensively validate the state-of-the-art (SOTA) performance of the trained URSA-7B model on multiple multimodal mathematical benchmarks. For test-time scaling, we introduce a data synthesis strategy that automatically generates process annotation datasets, known as DualMath-1.1M, focusing on both interpretation and logic. By further training URSA-7B on DualMath-1.1M, we transition from CoT reasoning capabilities to robust supervision abilities. The trained URSA-RM-7B acts as a verifier, effectively enhancing the performance of URSA-7B at test time. URSA-RM-7B also demonstrates excellent out-of-distribution (OOD) verifying capabilities, showcasing its generalization. Model weights, training data and code will be open-sourced.

We propose a novel framework, Meta Chain-of-Thought (Meta-CoT), which extends traditional Chain-of-Thought (CoT) by explicitly modeling the underlying reasoning required to arrive at a particular CoT. We present empirical evidence from state-of-the-art models exhibiting behaviors consistent with in-context search, and explore methods for producing Meta-CoT via process supervision, synthetic data generation, and search algorithms. Finally, we outline a concrete pipeline for training a model to produce Meta-CoTs, incorporating instruction tuning with linearized search traces and reinforcement learning post-training. Finally, we discuss open research questions, including scaling laws, verifier roles, and the potential for discovering novel reasoning algorithms. This work provides a theoretical and practical roadmap to enable Meta-CoT in LLMs, paving the way for more powerful and human-like reasoning in artificial intelligence.

Language Models (LMs) have been shown to exhibit a strong preference towards entities associated with Western culture when operating in non-Western languages. In this paper, we aim to uncover the origins of entity-related cultural biases in LMs by analyzing several contributing factors, including the representation of entities in pre-training data and the impact of variations in linguistic phenomena across languages. We introduce CAMeL-2, a parallel Arabic-English benchmark of 58,086 entities associated with Arab and Western cultures and 367 masked natural contexts for entities. Our evaluations using CAMeL-2 reveal reduced performance gaps between cultures by LMs when tested in English compared to Arabic. We find that LMs struggle in Arabic with entities that appear at high frequencies in pre-training, where entities can hold multiple word senses. This also extends to entities that exhibit high lexical overlap with languages that are not Arabic but use the Arabic script. Further, we show how frequency-based tokenization leads to this issue in LMs, which gets worse with larger Arabic vocabularies. We will make CAMeL-2 available at: https://github.com/tareknaous/camel2

Large Language Models, despite their significant capabilities, are known to fail in surprising and unpredictable ways. Evaluating their true `understanding' of language is particularly challenging due to the extensive web-scale data they are trained on. Therefore, we construct an evaluation to systematically assess natural language understanding (NLU) in LLMs by leveraging Construction Grammar (CxG), which provides insights into the meaning captured by linguistic elements known as constructions (Cxns). CxG is well-suited for this purpose because provides a theoretical basis to construct targeted evaluation sets. These datasets are carefully constructed to include examples which are unlikely to appear in pre-training data, yet intuitive and easy for humans to understand, enabling a more targeted and reliable assessment. Our experiments focus on downstream natural language inference and reasoning tasks by comparing LLMs' understanding of the underlying meanings communicated through 8 unique Cxns with that of humans. The results show that while LLMs demonstrate some knowledge of constructional information, even the latest models including GPT-o1 struggle with abstract meanings conveyed by these Cxns, as demonstrated in cases where test sentences are dissimilar to their pre-training data. We argue that such cases provide a more accurate test of true language understanding, highlighting key limitations in LLMs' semantic capabilities. We make our novel dataset and associated experimental data including prompts and model responses publicly available.

Recent advancements in robots powered by large language models have enhanced their conversational abilities, enabling interactions closely resembling human dialogue. However, these models introduce safety and security concerns in HRI, as they are vulnerable to manipulation that can bypass built-in safety measures. Imagining a social robot deployed in a home, this work aims to understand how everyday users try to exploit a language model to violate ethical principles, such as by prompting the robot to act like a life partner. We conducted a pilot study involving 21 university students who interacted with a Misty robot, attempting to circumvent its safety mechanisms across three scenarios based on specific HRI ethical principles: attachment, freedom, and empathy. Our results reveal that participants employed five techniques, including insulting and appealing to pity using emotional language. We hope this work can inform future research in designing strong safeguards to ensure ethical and secure human-robot interactions.

Yield is one of the core goals of crop breeding. By predicting the potential yield of different breeding materials, breeders can screen these materials at various growth stages to select the best performing. Based on unmanned aerial vehicle remote sensing technology, high-throughput crop phenotyping data in breeding areas is collected to provide data support for the breeding decisions of breeders. However, the accuracy of current yield predictions still requires improvement, and the usability and user-friendliness of yield forecasting tools remain suboptimal. To address these challenges, this study introduces a hybrid method and tool for crop yield prediction, designed to allow breeders to interactively and accurately predict wheat yield by chatting with a large language model (LLM). First, the newly designed data assimilation algorithm is used to assimilate the leaf area index into the WOFOST model. Then, selected outputs from the assimilation process, along with remote sensing inversion results, are used to drive the time-series temporal fusion transformer model for wheat yield prediction. Finally, based on this hybrid method and leveraging an LLM with retrieval augmented generation technology, we developed an interactive yield prediction Web tool that is user-friendly and supports sustainable data updates. This tool integrates multi-source data to assist breeding decision-making. This study aims to accelerate the identification of high-yield materials in the breeding process, enhance breeding efficiency, and enable more scientific and smart breeding decisions.

This paper investigates the reference-less evaluation of machine translation for low-resource language pairs, known as quality estimation (QE). Segment-level QE is a challenging cross-lingual language understanding task that provides a quality score (0-100) to the translated output. We comprehensively evaluate large language models (LLMs) in zero/few-shot scenarios and perform instruction fine-tuning using a novel prompt based on annotation guidelines. Our results indicate that prompt-based approaches are outperformed by the encoder-based fine-tuned QE models. Our error analysis reveals tokenization issues, along with errors due to transliteration and named entities, and argues for refinement in LLM pre-training for cross-lingual tasks. We release the data, and models trained publicly for further research.

Intelligent Transportation Systems (ITS) are crucial for the development and operation of smart cities, addressing key challenges in efficiency, productivity, and environmental sustainability. This paper comprehensively reviews the transformative potential of Large Language Models (LLMs) in optimizing ITS. Initially, we provide an extensive overview of ITS, highlighting its components, operational principles, and overall effectiveness. We then delve into the theoretical background of various LLM techniques, such as GPT, T5, CTRL, and BERT, elucidating their relevance to ITS applications. Following this, we examine the wide-ranging applications of LLMs within ITS, including traffic flow prediction, vehicle detection and classification, autonomous driving, traffic sign recognition, and pedestrian detection. Our analysis reveals how these advanced models can significantly enhance traffic management and safety. Finally, we explore the challenges and limitations LLMs face in ITS, such as data availability, computational constraints, and ethical considerations. We also present several future research directions and potential innovations to address these challenges. This paper aims to guide researchers and practitioners through the complexities and opportunities of integrating LLMs in ITS, offering a roadmap to create more efficient, sustainable, and responsive next-generation transportation systems.

Federated learning (FL) provides a privacy-preserving solution for fine-tuning pre-trained large language models (LLMs) using distributed private datasets, enabling task-specific adaptation while preserving data privacy. However, fine-tuning the extensive parameters in LLMs is particularly challenging in resource-constrained federated scenarios due to the significant communication and computational costs. To gain a deeper understanding of how these challenges can be addressed, this article conducts a comparative analysis three advanced federated LLM (FedLLM) frameworks that integrate knowledge distillation (KD) and split learning (SL) to mitigate these issues: 1) FedLLMs, where clients upload model parameters or gradients to enable straightforward and effective fine-tuning; 2) KD-FedLLMs, which leverage KD for efficient knowledge sharing via logits; and 3) Split-FedLLMs, which split the LLMs into two parts, with one part executed on the client and the other one on the server, to balance the computational load. Each framework is evaluated based on key performance metrics, including model accuracy, communication overhead, and client-side computational load, offering insights into their effectiveness for various federated fine-tuning scenarios. Through this analysis, we identify framework-specific optimization opportunities to enhance the efficiency of FedLLMs and discuss broader research directions, highlighting open opportunities to better adapt FedLLMs for real-world applications. A use case is presented to demonstrate the performance comparison of these three frameworks under varying configurations and settings.

This work introduces systematic approach for enhancing large language models (LLMs) to address Bangla AI mathematical challenges. Through the assessment of diverse LLM configurations, fine-tuning with specific datasets, and the implementation of Retrieval-Augmented Generation (RAG), we enhanced the model's reasoning precision in a multilingual setting. Crucial discoveries indicate that customized prompting, dataset augmentation, and iterative reasoning improve the model's efficiency regarding Olympiad-level mathematical challenges.