Traditional autonomous driving methods adopt a modular design, decomposing tasks into sub-tasks. In contrast, end-to-end autonomous driving directly outputs actions from raw sensor data, avoiding error accumulation. However, training an end-to-end model requires a comprehensive dataset; otherwise, the model exhibits poor generalization capabilities. Recently, large language models (LLMs) have been applied to enhance the generalization capabilities of end-to-end driving models. Most studies explore LLMs in an open-loop manner, where the output actions are compared to those of experts without direct feedback from the real world, while others examine closed-loop results only in simulations. This paper proposes an efficient architecture that integrates multimodal LLMs into end-to-end driving models operating in closed-loop settings in real-world environments. In our architecture, the LLM periodically processes raw sensor data to generate high-level driving instructions, effectively guiding the end-to-end model, even at a slower rate than the raw sensor data. This architecture relaxes the trade-off between the latency and inference quality of the LLM. It also allows us to choose from a wide variety of LLMs to improve high-level driving instructions and minimize fine-tuning costs. Consequently, our architecture reduces data collection requirements because the LLMs do not directly output actions; we only need to train a simple imitation learning model to output actions. In our experiments, the training data for the end-to-end model in a real-world environment consists of only simple obstacle configurations with one traffic cone, while the test environment is more complex and contains multiple obstacles placed in various positions. Experiments show that the proposed architecture enhances the generalization capabilities of the end-to-end model even without fine-tuning the LLM.

Generating large-scale, domain-specific, multilingual multi-turn dialogue datasets remains a significant hurdle for training effective Multi-Turn Intent Classification models in chatbot systems. In this paper, we introduce Chain-of-Intent, a novel mechanism that combines Hidden Markov Models with Large Language Models (LLMs) to generate contextually aware, intent-driven conversations through self-play. By extracting domain-specific knowledge from e-commerce chat logs, we estimate conversation turns and intent transitions, which guide the generation of coherent dialogues. Leveraging LLMs to enhance emission probabilities, our approach produces natural and contextually consistent questions and answers. We also propose MINT-CL, a framework for multi-turn intent classification using multi-task contrastive learning, improving classification accuracy without the need for extensive annotated data. Evaluations show that our methods outperform baselines in dialogue quality and intent classification accuracy, especially in multilingual settings, while significantly reducing data generation efforts. Furthermore, we release MINT-E, a multilingual, intent-aware multi-turn e-commerce dialogue corpus to support future research in this area.

LLM-based autonomous agents have demonstrated outstanding performance in solving complex industrial tasks. However, in the pursuit of carbon neutrality and high-performance renewable energy systems, existing AI-assisted design automation faces significant limitations in explainability, scalability, and usability. To address these challenges, we propose LP-COMDA, an LLM-based, physics-informed autonomous agent that automates the modulation design of power converters in Power Electronics Systems with minimal human supervision. Unlike traditional AI-assisted approaches, LP-COMDA contains an LLM-based planner that gathers and validates design specifications through a user-friendly chat interface. The planner then coordinates with physics-informed design and optimization tools to iteratively generate and refine modulation designs autonomously. Through the chat interface, LP-COMDA provides an explainable design process, presenting explanations and charts. Experiments show that LP-COMDA outperforms all baseline methods, achieving a 63.2% reduction in error compared to the second-best benchmark method in terms of standard mean absolute error. Furthermore, empirical studies with 20 experts conclude that design time with LP-COMDA is over 33 times faster than conventional methods, showing its significant improvement on design efficiency over the current processes.

Recent improvements in visual synthesis have significantly enhanced the depiction of generated human photos, which are pivotal due to their wide applicability and demand. Nonetheless, the existing text-to-image or text-to-video models often generate low-quality human photos that might differ considerably from real-world body structures, referred to as "abnormal human bodies". Such abnormalities, typically deemed unacceptable, pose considerable challenges in the detection and repair of them within human photos. These challenges require precise abnormality recognition capabilities, which entail pinpointing both the location and the abnormality type. Intuitively, Visual Language Models (VLMs) that have obtained remarkable performance on various visual tasks are quite suitable for this task. However, their performance on abnormality detection in human photos is quite poor. Hence, it is quite important to highlight this task for the research community. In this paper, we first introduce a simple yet challenging task, i.e., \textbf{F}ine-grained \textbf{H}uman-body \textbf{A}bnormality \textbf{D}etection \textbf{(FHAD)}, and construct two high-quality datasets for evaluation. Then, we propose a meticulous framework, named HumanCalibrator, which identifies and repairs abnormalities in human body structures while preserving the other content. Experiments indicate that our HumanCalibrator achieves high accuracy in abnormality detection and accomplishes an increase in visual comparisons while preserving the other visual content.

Scientific progress depends on researchers' ability to synthesize the growing body of literature. Can large language models (LMs) assist scientists in this task? We introduce OpenScholar, a specialized retrieval-augmented LM that answers scientific queries by identifying relevant passages from 45 million open-access papers and synthesizing citation-backed responses. To evaluate OpenScholar, we develop ScholarQABench, the first large-scale multi-domain benchmark for literature search, comprising 2,967 expert-written queries and 208 long-form answers across computer science, physics, neuroscience, and biomedicine. On ScholarQABench, OpenScholar-8B outperforms GPT-4o by 5% and PaperQA2 by 7% in correctness, despite being a smaller, open model. While GPT4o hallucinates citations 78 to 90% of the time, OpenScholar achieves citation accuracy on par with human experts. OpenScholar's datastore, retriever, and self-feedback inference loop also improves off-the-shelf LMs: for instance, OpenScholar-GPT4o improves GPT-4o's correctness by 12%. In human evaluations, experts preferred OpenScholar-8B and OpenScholar-GPT4o responses over expert-written ones 51% and 70% of the time, respectively, compared to GPT4o's 32%. We open-source all of our code, models, datastore, data and a public demo.

Language models perform differently across languages. It has been previously suggested that morphological typology may explain some of this variability (Cotterell et al., 2018). We replicate previous analyses and find additional new evidence for a performance gap between agglutinative and fusional languages, where fusional languages, such as English, tend to have better language modeling performance than morphologically more complex languages like Turkish. We then propose and test three possible causes for this performance gap: morphological alignment of tokenizers, tokenization quality, and disparities in dataset sizes and measurement. To test the morphological alignment hypothesis, we present MorphScore, a tokenizer evaluation metric, and supporting datasets for 22 languages. We find some evidence that tokenization quality explains the performance gap, but none for the role of morphological alignment. Instead we find that the performance gap is most reduced when training datasets are of equivalent size across language types, but only when scaled according to the so-called "byte-premium" -- the different encoding efficiencies of different languages and orthographies. These results suggest that no language is harder or easier for a language model to learn on the basis of its morphological typology. Differences in performance can be attributed to disparities in dataset size. These results bear on ongoing efforts to improve performance for low-performing and under-resourced languages.

Large Vision-Language Models (LVLMs) represent a significant advancement toward achieving superior multimodal capabilities by enabling powerful Large Language Models (LLMs) to understand visual input. Typically, LVLMs utilize visual encoders, such as CLIP, to transform images into visual tokens, which are then aligned with textual tokens through projection layers before being input into the LLM for inference. Although existing LVLMs have achieved significant success, their inference efficiency is still limited by the substantial number of visual tokens and the potential redundancy among them. To mitigate this issue, we propose Focal Pruning (FoPru), a training-free method that prunes visual tokens based on the attention-based token significance derived from the vision encoder. Specifically, we introduce two alternative pruning strategies: 1) the rank strategy, which leverages all token significance scores to retain more critical tokens in a global view; 2) the row strategy, which focuses on preserving continuous key information in images from a local perspective. Finally, the selected tokens are reordered to maintain their original positional relationships. Extensive experiments across various LVLMs and multimodal datasets demonstrate that our method can prune a large number of redundant tokens while maintaining high accuracy, leading to significant improvements in inference efficiency.

The ability to perform complex reasoning across multimodal inputs is essential for models to effectively interact with humans in real-world scenarios. Advancements in vision-language models have significantly improved performance on tasks that require processing explicit and direct textual inputs, such as Visual Question Answering (VQA) and Visual Grounding (VG). However, less attention has been given to improving the model capabilities to comprehend nuanced and ambiguous forms of communication. This presents a critical challenge, as human language in real-world interactions often convey hidden intentions that rely on context for accurate interpretation. To address this gap, we propose VAGUE, a multimodal benchmark comprising 3.9K indirect human utterances paired with corresponding scenes. Additionally, we contribute a model-based pipeline for generating prompt-solution pairs from input images. Our work aims to delve deeper into the ability of models to understand indirect communication and seek to contribute to the development of models capable of more refined and human-like interactions. Extensive evaluation on multiple VLMs reveals that mainstream models still struggle with indirect communication when required to perform complex linguistic and visual reasoning. We release our code and data at https://github.com/Hazel-Heejeong-Nam/VAGUE.git.

Large Language Models (LLMs) have shown impressive proficiency across a range of natural language processing tasks yet remain vulnerable to adversarial prompts, known as jailbreak attacks, carefully designed to elicit harmful responses from LLMs. Traditional methods rely on manual heuristics, which suffer from limited generalizability. While being automatic, optimization-based attacks often produce unnatural jailbreak prompts that are easy to detect by safety filters or require high computational overhead due to discrete token optimization. Witnessing the limitations of existing jailbreak methods, we introduce Generative Adversarial Suffix Prompter (GASP), a novel framework that combines human-readable prompt generation with Latent Bayesian Optimization (LBO) to improve adversarial suffix creation in a fully black-box setting. GASP leverages LBO to craft adversarial suffixes by efficiently exploring continuous embedding spaces, gradually optimizing the model to improve attack efficacy while balancing prompt coherence through a targeted iterative refinement procedure. Our experiments show that GASP can generate natural jailbreak prompts, significantly improving attack success rates, reducing training times, and accelerating inference speed, thus making it an efficient and scalable solution for red-teaming LLMs.

Constructing high-quality Supervised Fine-Tuning (SFT) datasets is critical for the training of large language models (LLMs). Recent studies have shown that using data from a specific source, Ruozhiba, a Chinese website where users ask "silly" questions to better understand certain topics, can lead to better fine-tuning performance. This paper aims to explore some hidden factors: the potential interpretations of its success and a large-scale evaluation of the performance. First, we leverage GPT-4 to analyze the successful cases of Ruozhiba questions from the perspective of education, psychology, and cognitive science, deriving a set of explanatory rules. Then, we construct fine-tuning datasets by applying these rules to the MMLU training set. Surprisingly, our results indicate that rules can significantly improve model performance in certain tasks, while potentially diminishing performance on others. For example, SFT data generated following the "Counterintuitive Thinking" rule can achieve approximately a 5% improvement on the "Global Facts" task, whereas the "Blurring the Conceptual Boundaries" rule leads to a performance drop of 6.14% on the "Econometrics" task. In addition, for specific tasks, different rules tend to have a consistent impact on model performance. This suggests that the differences between the extracted rules are not as significant, and the effectiveness of the rules is relatively consistent across tasks. Our research highlights the importance of considering task diversity and rule applicability when constructing SFT datasets to achieve more comprehensive performance improvements.