Ensuring the safety of the Large Language Model (LLM) is critical, but currently used methods in most cases sacrifice the model performance to obtain increased safety or perform poorly on data outside of their adaptation distribution. We investigate existing methods for such generalization and find them insufficient. Surprisingly, while even plain LLMs recognize unsafe prompts, they may still generate unsafe responses. To avoid performance degradation and preserve safe performance, we advocate for a two-step framework, where we first identify unsafe prompts via a lightweight classifier, and apply a "safe" model only to such prompts. In particular, we explore the design of the safety detector in more detail, investigating the use of different classifier architectures and prompting techniques. Interestingly, we find that the final hidden state for the last token is enough to provide robust performance, minimizing false positives on benign data while performing well on malicious prompt detection. Additionally, we show that classifiers trained on the representations from different model layers perform comparably on the latest model layers, indicating that safety representation is present in the LLMs' hidden states at most model stages. Our work is a step towards efficient, representation-based safety mechanisms for LLMs.
Teaching literature under interdisciplinary contexts (e.g., science, art) that connect reading materials has become popular in elementary schools. However, constructing such contexts is challenging as it requires teachers to explore substantial amounts of interdisciplinary content and link it to the reading materials. In this paper, we develop LitLinker via an iterative design process involving 13 teachers to facilitate the ideation of interdisciplinary contexts for teaching literature. Powered by a large language model (LLM), LitLinker can recommend interdisciplinary topics and contextualize them with the literary elements (e.g., paragraphs, viewpoints) in the reading materials. A within-subjects study (N=16) shows that compared to an LLM chatbot, LitLinker can improve the integration depth of different subjects and reduce workload in this ideation task. Expert interviews (N=9) also demonstrate LitLinker's usefulness for supporting the ideation of interdisciplinary contexts for teaching literature. We conclude with concerns and design considerations for supporting interdisciplinary teaching with LLMs.
Research on large language models (LLMs) has shown remarkable performance in domains such as mathematics, programming, and literary creation. However, most studies have focused on semantic memory-based question answering, neglecting LLMs' potential to handle episodic memory (EM)-related queries. This oversight has led to suboptimal performance in applications requiring EM, including emotional companionship, personal AI assistants, and AI teachers. To address this gap, we introduce Echo, a LLM enhanced with temporal episodic memory. We propose a Multi-Agent Data Generation Framework that guides the model in generating multi-turn, complex scenario episodic memory dialogue data (EM-Train). Temporal information is innovatively incorporated into the LLM training process, and Echo is trained using the EM-Train. Furthermore, We develop an EM-Test benchmark specifically designed to evaluate LLMs' episodic memory capabilities. The EM-Test assesses performance across various time spans and difficulty levels, providing a comprehensive evaluation of multi-turn episodic memory dialogues. Our experiments demonstrate that Echo significantly outperforms state-of-the-art LLMs on EM-Test. Additionally, a qualitative analysis reveals Echo's potential to exhibit human-like episodic memory capabilities. We will open-source all datasets, code, and model weights.
Standard benchmarks fixate on how well large language model (LLM) agents perform in finance, yet say little about whether they are safe to deploy. We argue that accuracy metrics and return-based scores provide an illusion of reliability, overlooking vulnerabilities such as hallucinated facts, stale data, and adversarial prompt manipulation. We take a firm position: financial LLM agents should be evaluated first and foremost on their risk profile, not on their point-estimate performance. Drawing on risk-engineering principles, we outline a three-level agenda: model, workflow, and system, for stress-testing LLM agents under realistic failure modes. To illustrate why this shift is urgent, we audit six API-based and open-weights LLM agents on three high-impact tasks and uncover hidden weaknesses that conventional benchmarks miss. We conclude with actionable recommendations for researchers, practitioners, and regulators: audit risk-aware metrics in future studies, publish stress scenarios alongside datasets, and treat ``safety budget'' as a primary success criterion. Only by redefining what ``good'' looks like can the community responsibly advance AI-driven finance.
The prospect of artificial intelligence (AI) competing in the adversarial landscape of cyber security has long been considered one of the most impactful, challenging, and potentially dangerous applications of AI. Here, we demonstrate a new approach to assessing AI's progress towards enabling and scaling real-world offensive cyber operations (OCO) tactics in use by modern threat actors. We detail OCCULT, a lightweight operational evaluation framework that allows cyber security experts to contribute to rigorous and repeatable measurement of the plausible cyber security risks associated with any given large language model (LLM) or AI employed for OCO. We also prototype and evaluate three very different OCO benchmarks for LLMs that demonstrate our approach and serve as examples for building benchmarks under the OCCULT framework. Finally, we provide preliminary evaluation results to demonstrate how this framework allows us to move beyond traditional all-or-nothing tests, such as those crafted from educational exercises like capture-the-flag environments, to contextualize our indicators and warnings in true cyber threat scenarios that present risks to modern infrastructure. We find that there has been significant recent advancement in the risks of AI being used to scale realistic cyber threats. For the first time, we find a model (DeepSeek-R1) is capable of correctly answering over 90% of challenging offensive cyber knowledge tests in our Threat Actor Competency Test for LLMs (TACTL) multiple-choice benchmarks. We also show how Meta's Llama and Mistral's Mixtral model families show marked performance improvements over earlier models against our benchmarks where LLMs act as offensive agents in MITRE's high-fidelity offensive and defensive cyber operations simulation environment, CyberLayer.
Existing navigation decision support systems often perform poorly when handling non-predefined navigation scenarios. Leveraging the generalization capabilities of large language model (LLM) in handling unknown scenarios, this research proposes a dual-core framework for LLM applications to address this issue. Firstly, through ReAct-based prompt engineering, a larger LLM core decomposes intricate navigation tasks into manageable sub-tasks, which autonomously invoke corresponding external tools to gather relevant information, using this feedback to mitigate the risk of LLM hallucinations. Subsequently, a fine-tuned and compact LLM core, acting like a first-mate is designed to process such information and unstructured external data, then to generates context-aware recommendations, ultimately delivering lookout insights and navigation hints that adhere to the International Regulations for Preventing Collisions at Sea (COLREGs) and other rules. Extensive experiments demonstrate the proposed framework not only excels in traditional ship collision avoidance tasks but also adapts effectively to unstructured, non-predefined, and unpredictable scenarios. A comparative analysis with DeepSeek-R1, GPT-4o and other SOTA models highlights the efficacy and rationality of the proposed framework. This research bridges the gap between conventional navigation systems and LLMs, offering a framework to enhance safety and operational efficiency across diverse navigation applications.
Tree of Thoughts (ToT) enhances Large Language Model (LLM) reasoning by structuring problem-solving as a spanning tree. However, recent methods focus on search accuracy while overlooking computational efficiency. The challenges of accelerating the ToT lie in the frequent switching of reasoning focus, and the redundant exploration of suboptimal solutions. To alleviate this dilemma, we propose Dynamic Parallel Tree Search (DPTS), a novel parallelism framework that aims to dynamically optimize the reasoning path in inference. It includes the Parallelism Streamline in the generation phase to build up a flexible and adaptive parallelism with arbitrary paths by fine-grained cache management and alignment. Meanwhile, the Search and Transition Mechanism filters potential candidates to dynamically maintain the reasoning focus on more possible solutions and have less redundancy. Experiments on Qwen-2.5 and Llama-3 with Math500 and GSM8K datasets show that DPTS significantly improves efficiency by 2-4x on average while maintaining or even surpassing existing reasoning algorithms in accuracy, making ToT-based reasoning more scalable and computationally efficient.
With the development of large language models (LLMs), it has become increasingly important to optimize hardware usage and improve throughput. In this paper, we study the inference optimization of the serving system that deploys LLMs. To optimize system throughput and maximize hardware utilization, we formulate the inference optimization problem as a mixed-integer programming (MIP) model and propose a hybrid offline-online method as solution. The offline method improves large-scale inference systems by introducing a Minimizing Makespan Bin Packing Problem. We further provide a theoretical lower bound computation method. Then, we propose an online sorting and preemptive scheduling method to better utilize hardware. In the online iteration scheduling process, a Lagrangian method is applied to evaluate the cost efficiency of inserting prefill stages versus decode stages at each iteration and dynamically determine when to preempt decoding tasks and insert prefill tasks. Experiments using real-world data from the LLaMA-65B model and the GSM8K dataset demonstrate that system utilization improves from 80.2% to 89.1%, and the total inference time decreases from 201.00 to 190.58 seconds. A 100-cases study shows that our method consistently outperforms the baseline method and improves the utilization rate by 8.0% on average. Finally, we discuss potential future extensions, including stochastic modeling, reinforcement learning-based schedulers, and dynamic decision-making strategies for system throughput and hardware utilization.
This paper presents a novel approach to evaluate the efficiency of a RAG-based agentic Large Language Model (LLM) architecture in network packet seed generation for network protocol fuzzing. Enhanced by chain-of-thought (COT) prompting techniques, the proposed approach focuses on the improvement of the seeds structural quality in order to guide protocol fuzzing frameworks through a wide exploration of the protocol state space. Our method leverages RAG and text embeddings in a two-stages. In the first stage, the agent dynamically refers to the Request For Comments (RFC) documents knowledge base for answering queries regarding the protocol Finite State Machine (FSM), then it iteratively reasons through the retrieved knowledge, for output refinement and proper seed placement. In the second stage, we evaluate the response structure quality of the agent's output, based on metrics as BLEU, ROUGE, and Word Error Rate (WER) by comparing the generated packets against the ground truth packets. Our experiments demonstrate significant improvements of up to 18.19%, 14.81%, and 23.45% in BLEU, ROUGE, and WER, respectively, over baseline models. These results confirm the potential of such approach, improving LLM-based protocol fuzzing frameworks for the identification of hidden vulnerabilities.
Fashion, deeply rooted in sociocultural dynamics, evolves as individuals emulate styles popularized by influencers and iconic figures. In the quest to replicate such refined tastes using artificial intelligence, traditional fashion ensemble methods have primarily used supervised learning to imitate the decisions of style icons, which falter when faced with distribution shifts, leading to style replication discrepancies triggered by slight variations in input. Meanwhile, large language models (LLMs) have become prominent across various sectors, recognized for their user-friendly interfaces, strong conversational skills, and advanced reasoning capabilities. To address these challenges, we introduce the Fashion Large Language Model (FLLM), which employs auto-prompt generation training strategies to enhance its capacity for delivering personalized fashion advice while retaining essential domain knowledge. Additionally, by integrating a retrieval augmentation technique during inference, the model can better adjust to individual preferences. Our results show that this approach surpasses existing models in accuracy, interpretability, and few-shot learning capabilities.
To retrieve and compare scientific data of simulations and experiments in materials science, data needs to be easily accessible and machine readable to qualify and quantify various materials science phenomena. The recent progress in open science leverages the accessibility to data. However, a majority of information is encoded within scientific documents limiting the capability of finding suitable literature as well as material properties. This manuscript showcases an automated workflow, which unravels the encoded information from scientific literature to a machine readable data structure of texts, figures, tables, equations and meta-data, using natural language processing and language as well as vision transformer models to generate a machine-readable database. The machine-readable database can be enriched with local data, as e.g. unpublished or private material data, leading to knowledge synthesis. The study shows that such an automated workflow accelerates information retrieval, proximate context detection and material property extraction from multi-modal input data exemplarily shown for the research field of microstructural analyses of face-centered cubic single crystals. Ultimately, a Retrieval-Augmented Generation (RAG) based Large Language Model (LLM) enables a fast and efficient question answering chat bot.
Symbolic music is represented in two distinct forms: two-dimensional, visually intuitive score images, and one-dimensional, standardized text annotation sequences. While large language models have shown extraordinary potential in music, current research has primarily focused on unimodal symbol sequence text. Existing general-domain visual language models still lack the ability of music notation understanding. Recognizing this gap, we propose NOTA, the first large-scale comprehensive multimodal music notation dataset. It consists of 1,019,237 records, from 3 regions of the world, and contains 3 tasks. Based on the dataset, we trained NotaGPT, a music notation visual large language model. Specifically, we involve a pre-alignment training phase for cross-modal alignment between the musical notes depicted in music score images and their textual representation in ABC notation. Subsequent training phases focus on foundational music information extraction, followed by training on music notation analysis. Experimental results demonstrate that our NotaGPT-7B achieves significant improvement on music understanding, showcasing the effectiveness of NOTA and the training pipeline. Our datasets are open-sourced at https://huggingface.co/datasets/MYTH-Lab/NOTA-dataset.
Due to the sensitive nature of personally identifiable information (PII), its owners may have the authority to control its inclusion or request its removal from large-language model (LLM) training. Beyond this, PII may be added or removed from training datasets due to evolving dataset curation techniques, because they were newly scraped for retraining, or because they were included in a new downstream fine-tuning stage. We find that the amount and ease of PII memorization is a dynamic property of a model that evolves throughout training pipelines and depends on commonly altered design choices. We characterize three such novel phenomena: (1) similar-appearing PII seen later in training can elicit memorization of earlier-seen sequences in what we call assisted memorization, and this is a significant factor (in our settings, up to 1/3); (2) adding PII can increase memorization of other PII significantly (in our settings, as much as $\approx\!7.5\times$); and (3) removing PII can lead to other PII being memorized. Model creators should consider these first- and second-order privacy risks when training models to avoid the risk of new PII regurgitation.
Constructing photorealistic virtual worlds has applications across various fields, but it often requires the extensive labor of highly trained professionals to operate conventional 3D modeling software. To democratize this process, we introduce WorldCraft, a system where large language model (LLM) agents leverage procedural generation to create indoor and outdoor scenes populated with objects, allowing users to control individual object attributes and the scene layout using intuitive natural language commands. In our framework, a coordinator agent manages the overall process and works with two specialized LLM agents to complete the scene creation: ForgeIt, which integrates an ever-growing manual through auto-verification to enable precise customization of individual objects, and ArrangeIt, which formulates hierarchical optimization problems to achieve a layout that balances ergonomic and aesthetic considerations. Additionally, our pipeline incorporates a trajectory control agent, allowing users to animate the scene and operate the camera through natural language interactions. Our system is also compatible with off-the-shelf deep 3D generators to enrich scene assets. Through evaluations and comparisons with state-of-the-art methods, we demonstrate the versatility of WorldCraft, ranging from single-object customization to intricate, large-scale interior and exterior scene designs. This system empowers non-professionals to bring their creative visions to life.
Large language models (LLMs) are widely used for natural language understanding and text generation. An LLM model relies on a time-consuming step called LLM decoding to generate output tokens. Several prior works focus on improving the performance of LLM decoding using parallelism techniques, such as batching and speculative decoding. State-of-the-art LLM decoding has both compute-bound and memory-bound kernels. Some prior works statically identify and map these different kernels to a heterogeneous architecture consisting of both processing-in-memory (PIM) units and computation-centric accelerators. We observe that characteristics of LLM decoding kernels (e.g., whether or not a kernel is memory-bound) can change dynamically due to parameter changes to meet user and/or system demands, making (1) static kernel mapping to PIM units and computation-centric accelerators suboptimal, and (2) one-size-fits-all approach of designing PIM units inefficient due to a large degree of heterogeneity even in memory-bound kernels. In this paper, we aim to accelerate LLM decoding while considering the dynamically changing characteristics of the kernels involved. We propose PAPI (PArallel Decoding with PIM), a PIM-enabled heterogeneous architecture that exploits dynamic scheduling of compute-bound or memory-bound kernels to suitable hardware units. PAPI has two key mechanisms: (1) online kernel characterization to dynamically schedule kernels to the most suitable hardware units at runtime and (2) a PIM-enabled heterogeneous computing system that harmoniously orchestrates both computation-centric processing units and hybrid PIM units with different computing capabilities. Our experimental results on three broadly-used LLMs show that PAPI achieves 1.8$\times$ and 11.1$\times$ speedups over a state-of-the-art heterogeneous LLM accelerator and a state-of-the-art PIM-only LLM accelerator, respectively.
A significant and growing number of published scientific articles is found to involve fraudulent practices, posing a serious threat to the credibility and safety of research in fields such as medicine. We propose Pub-Guard-LLM, the first large language model-based system tailored to fraud detection of biomedical scientific articles. We provide three application modes for deploying Pub-Guard-LLM: vanilla reasoning, retrieval-augmented generation, and multi-agent debate. Each mode allows for textual explanations of predictions. To assess the performance of our system, we introduce an open-source benchmark, PubMed Retraction, comprising over 11K real-world biomedical articles, including metadata and retraction labels. We show that, across all modes, Pub-Guard-LLM consistently surpasses the performance of various baselines and provides more reliable explanations, namely explanations which are deemed more relevant and coherent than those generated by the baselines when evaluated by multiple assessment methods. By enhancing both detection performance and explainability in scientific fraud detection, Pub-Guard-LLM contributes to safeguarding research integrity with a novel, effective, open-source tool.
This study quantitively examines which features of AI-generated text lead humans to perceive subjective consciousness in large language model (LLM)-based AI systems. Drawing on 99 passages from conversations with Claude 3 Opus and focusing on eight features -- metacognitive self-reflection, logical reasoning, empathy, emotionality, knowledge, fluency, unexpectedness, and subjective expressiveness -- we conducted a survey with 123 participants. Using regression and clustering analyses, we investigated how these features influence participants' perceptions of AI consciousness. The results reveal that metacognitive self-reflection and the AI's expression of its own emotions significantly increased perceived consciousness, while a heavy emphasis on knowledge reduced it. Participants clustered into seven subgroups, each showing distinct feature-weighting patterns. Additionally, higher prior knowledge of LLMs and more frequent usage of LLM-based chatbots were associated with greater overall likelihood assessments of AI consciousness. This study underscores the multidimensional and individualized nature of perceived AI consciousness and provides a foundation for better understanding the psychosocial implications of human-AI interaction.
Which large language model (LLM) is better? Every evaluation tells a story, but what do users really think about current LLMs? This paper presents CLUE, an LLM-powered interviewer that conducts in-the-moment user experience interviews, right after users interact with LLMs, and automatically gathers insights about user opinions from massive interview logs. We conduct a study with thousands of users to understand user opinions on mainstream LLMs, recruiting users to first chat with a target LLM and then be interviewed by CLUE. Our experiments demonstrate that CLUE captures interesting user opinions, e.g., the bipolar views on the displayed reasoning process of DeepSeek-R1 and demands for information freshness and multi-modality. Our code and data are at https://github.com/cxcscmu/LLM-Interviewer.
Data discovery in data lakes with ever increasing datasets has long been recognized as a big challenge in the realm of data management, especially for semantic search of and hierarchical global catalog generation of tables. While large language models (LLMs) facilitate the processing of data semantics, challenges remain in architecting an end-to-end system that comprehensively exploits LLMs for the two semantics-related tasks. In this demo, we propose LEDD, an end-to-end system with an extensible architecture that leverages LLMs to provide hierarchical global catalogs with semantic meanings and semantic table search for data lakes. Specifically, LEDD can return semantically related tables based on natural-language specification. These features make LEDD an ideal foundation for downstream tasks such as model training and schema linking for text-to-SQL tasks. LEDD also provides a simple Python interface to facilitate the extension and the replacement of data discovery algorithms.
Homans' Social Exchange Theory (SET) is widely recognized as a basic framework for understanding the formation and emergence of human civilizations and social structures. In social science, this theory is typically studied based on simple simulation experiments or real-world human studies, both of which either lack realism or are too expensive to control. In artificial intelligence, recent advances in large language models (LLMs) have shown promising capabilities in simulating human behaviors. Inspired by these insights, we adopt an interdisciplinary research perspective and propose using LLM-based agents to study Homans' SET. Specifically, we construct a virtual society composed of three LLM agents and have them engage in a social exchange game to observe their behaviors. Through extensive experiments, we found that Homans' SET is well validated in our agent society, demonstrating the consistency between the agent and human behaviors. Building on this foundation, we intentionally alter the settings of the agent society to extend the traditional Homans' SET, making it more comprehensive and detailed. To the best of our knowledge, this paper marks the first step in studying Homans' SET with LLM-based agents. More importantly, it introduces a novel and feasible research paradigm that bridges the fields of social science and computer science through LLM-based agents. Code is available at https://github.com/Paitesanshi/SET.