High-quality training data has proven crucial for developing performant large language models (LLMs). However, commercial LLM providers disclose few, if any, details about the data used for training. This lack of transparency creates multiple challenges: it limits external oversight and inspection of LLMs for issues such as copyright infringement, it undermines the agency of data authors, and it hinders scientific research on critical issues such as data contamination and data selection. How can we recover what training data is known to LLMs? In this work, we demonstrate a new method to identify training data known to proprietary LLMs like GPT-4 without requiring any access to model weights or token probabilities, by using information-guided probes. Our work builds on a key observation: text passages with high surprisal are good search material for memorization probes. By evaluating a model's ability to successfully reconstruct high-surprisal tokens in text, we can identify a surprising number of texts memorized by LLMs.

The latest research on Large Language Models (LLMs) has demonstrated significant advancement in the field of Natural Language Processing (NLP). However, despite this progress, there is still a lack of reliability in these models. This is due to the stochastic architecture of LLMs, which presents a challenge for users attempting to ascertain the reliability of a model's response. These responses may cause serious harm in high-risk environments or expensive failures in industrial contexts. Therefore, we introduce the framework REpeated Clustering of Scores Improving the Precision (RECSIP) which focuses on improving the precision of LLMs by asking multiple models in parallel, scoring and clustering their responses to ensure a higher reliability on the response. The evaluation of our reference implementation recsip on the benchmark MMLU-Pro using the models GPT-4o, Claude and Gemini shows an overall increase of 5.8 per cent points compared to the best used model.

Large Language Models (LLMs) have made significant strides in generating human-like responses, largely due to preference alignment techniques. However, these methods often assume unbiased human feedback, which is rarely the case in real-world scenarios. This paper introduces Content-Aware Noise-Resilient Preference Optimization (CNRPO), a novel framework that addresses multiple sources of content-dependent noise in preference learning. CNRPO employs a multi-objective optimization approach to separate true preferences from contentaware noises, effectively mitigating their impact. We leverage backdoor attack mechanisms to efficiently learn and control various noise sources within a single model. Theoretical analysis and extensive experiments on different synthetic noisy datasets demonstrate that CNRPO significantly improves alignment with primary human preferences while controlling for secondary noises and biases, such as response length and harmfulness. Recent advancements in Large Language Models (LLMs) have significantly enhanced their ability through preference alignment techniques, primarily using Reinforcement Learning from Human Feedback (RLHF) (Christiano et al., 2023; Stiennon et al., 2022; Ouyang et al., 2022). However, RLHF faces challenges like reward model misgeneralization and training instability (Touvron et al., 2023; Casper et al., 2023). To address these issues, ranking-based methods like Direct Preference Optimization (DPO) (Rafailov et al., 2024) and Identity Preference Optimization (IPO) (Azar et al., 2023) have been developed, bypassing explicit reward modeling. While these approaches have advanced LLM capabilities, they assume unbiased human feedback. In reality, annotations can be influenced by various biases, such as a preference for longer responses or a focus on safety (Park et al., 2024b; Wang et al., 2024).

Fully Homomorphic Encryption over the torus (TFHE) enables computation on encrypted data without decryption, making it a cornerstone of secure and confidential computing. Despite its potential in privacy preserving machine learning, secure multi party computation, private blockchain transactions, and secure medical diagnostics, its adoption remains limited due to cryptographic complexity and usability challenges. While various TFHE libraries and compilers exist, practical code generation remains a hurdle. We propose a compiler integrated framework to evaluate LLM inference and agentic optimization for TFHE code generation, focusing on logic gates and ReLU activation. Our methodology assesses error rates, compilability, and structural similarity across open and closedsource LLMs. Results highlight significant limitations in off-the-shelf models, while agentic optimizations such as retrieval augmented generation (RAG) and few-shot prompting reduce errors and enhance code fidelity. This work establishes the first benchmark for TFHE code generation, demonstrating how LLMs, when augmented with domain-specific feedback, can bridge the expertise gap in FHE code generation.

We propose a novel dual-loop system that synergistically combines responsive neurostimulation (RNS) implants with artificial intelligence-driven wearable devices for treating post-traumatic stress disorder (PTSD) and enabling naturalistic brain research. In PTSD Therapy Mode, an implanted closed-loop neural device monitors amygdala activity and provides on-demand stimulation upon detecting pathological theta oscillations, while an ensemble of wearables (smart glasses, smartwatches, smartphones) uses multimodal large language model (LLM) analysis of sensory data to detect environmental or physiological PTSD triggers and deliver timely audiovisual interventions. Logged events from both the neural and wearable loops are analyzed to personalize trigger detection and progressively transition patients to non-invasive interventions. In Neuroscience Research Mode, the same platform is adapted for real-world brain activity capture. Wearable-LLM systems recognize naturalistic events (social interactions, emotional situations, compulsive behaviors, decision making) and signal implanted RNS devices (via wireless triggers) to record synchronized intracranial data during these moments. This approach builds on recent advances in mobile intracranial EEG recording and closed-loop neuromodulation in humans (BRAIN Initiative, 2023) (Mobbs et al., 2021). We discuss how our interdisciplinary system could revolutionize PTSD therapy and cognitive neuroscience by enabling 24/7 monitoring, context-aware intervention, and rich data collection outside traditional labs. The vision is a future where AI-enhanced devices continuously collaborate with the human brain, offering therapeutic support and deep insights into neural function, with the resulting real-world context rich neural data, in turn, accelerating the development of more biologically-grounded and human-centric AI.

Large language models (LLMs) are among the best methods for processing natural language, partly due to their versatility. At the same time, domain-specific LLMs are more practical in real-life applications. This work introduces a novel natural language dataset created by acquired data from official legislative authorities' websites. The study focuses on formulating three natural language processing (NLP) tasks to evaluate the effectiveness of LLMs on legislative content analysis within the context of the Polish legal system. Key findings highlight the potential of LLMs in automating and enhancing legislative content analysis while emphasizing specific challenges, such as understanding legal context. The research contributes to the advancement of NLP in the legal field, particularly in the Polish language. It has been demonstrated that even commonly accessible data can be practically utilized for legislative content analysis.

Large Language Models (LLMs) are increasingly being used to automate programming tasks. Yet, LLMs' capabilities in reasoning about program semantics are still inadequately studied, leaving significant potential for further exploration. This paper introduces FormalBench, a comprehensive benchmark designed to evaluate LLMs' reasoning abilities on program semantics, particularly via the task of synthesizing formal program specifications to assist verifying program correctness. This task requires both comprehensive reasoning over all possible program executions and the generation of precise, syntactically correct expressions that adhere to formal syntax and semantics. Using this benchmark, we evaluated the ability of LLMs in synthesizing consistent and complete specifications. Our findings show that LLMs perform well with simple control flows but struggle with more complex structures, especially loops, even with advanced prompting. Additionally, LLMs exhibit limited robustness against semantic-preserving transformations. We also highlight common failure patterns and design self-repair prompts, improving success rates by 25%.

The Unified Modeling Language is a standardized visual language widely used for modeling and documenting the design of software systems. Although many tools generate UML diagrams from UML code, generating executable UML code from image-based UML diagrams remains challenging. This paper proposes a new approach to generate UML code using a large multimodal language model automatically. Synthetic UML activity and sequence diagram datasets were created to train and test the model. We compared standard fine-tuning with LoRA techniques to optimize base models. The experiments measured code generation accuracy across different model sizes and training strategies. These results demonstrated that domain-adapted MM-LLMs perform for UML code generation automation, whereby, at the best model, it achieved BLEU and SSIM scores of 0.779 and 0.942 on sequence diagrams. This will enable the modernization of legacy systems and decrease the manual effort in software development workflows.

Advanced table question answering (TableQA) methods prompt large language models (LLMs) to generate answer text, SQL query, Python code, or custom operations, which impressively improve the complex reasoning problems in the TableQA task. However, these methods lack the versatility to cope with specific question types or table structures. In contrast, the Spreadsheet Formula, the widely-used and well-defined operation language for tabular data, has not been thoroughly explored to solve TableQA. In this paper, we first attempt to use Formula as the logical form for solving complex reasoning on the tables with different structures. Specifically, we construct a large Formula-annotated TableQA dataset \texttt{FromulaQA} from existing datasets. In addition, we propose \texttt{TabAF}, a general table answering framework to solve multiple types of tasks over multiple types of tables simultaneously. Unlike existing methods, \texttt{TabAF} decodes answers and Formulas with a single LLM backbone, demonstrating great versatility and generalization. \texttt{TabAF} based on Llama3.1-70B achieves new state-of-the-art performance on the WikiTableQuestion, HiTab and TabFact.

The demand for property valuation has attracted significant attention from sellers, buyers, and customers applying for loans. Reviews of existing approaches have revealed shortcomings in terms of not being able to handle missing value situations, as well as lacking interpretability, which means they cannot be used in real-world applications. To address these challenges, we propose an LLM-Generated EXplainable PRopErty valuation SyStem with neighbor imputation called EXPRESS, which provides the customizable missing value imputation technique, and addresses the opaqueness of prediction by providing the feature-wise explanation generated by LLM. The dynamic nearest neighbor search finds similar properties depending on different application scenarios by property configuration set by users (e.g., house age as criteria for the house in rural areas, and locations for buildings in urban areas). Motivated by the human appraisal procedure, we generate feature-wise explanations to provide users with a more intuitive understanding of the prediction results.