Source code authorship attribution is important in software forensics, plagiarism detection, and protecting software patch integrity. Existing techniques often rely on supervised machine learning, which struggles with generalization across different programming languages and coding styles due to the need for large labeled datasets. Inspired by recent advances in natural language authorship analysis using large language models (LLMs), which have shown exceptional performance without task-specific tuning, this paper explores the use of LLMs for source code authorship attribution. We present a comprehensive study demonstrating that state-of-the-art LLMs can successfully attribute source code authorship across different languages. LLMs can determine whether two code snippets are written by the same author with zero-shot prompting, achieving a Matthews Correlation Coefficient (MCC) of 0.78, and can attribute code authorship from a small set of reference code snippets via few-shot learning, achieving MCC of 0.77. Additionally, LLMs show some adversarial robustness against misattribution attacks. Despite these capabilities, we found that naive prompting of LLMs does not scale well with a large number of authors due to input token limitations. To address this, we propose a tournament-style approach for large-scale attribution. Evaluating this approach on datasets of C++ (500 authors, 26,355 samples) and Java (686 authors, 55,267 samples) code from GitHub, we achieve classification accuracy of up to 65% for C++ and 68.7% for Java using only one reference per author. These results open new possibilities for applying LLMs to code authorship attribution in cybersecurity and software engineering.

Models trained specifically to generate long Chains of Thought (CoTs) have recently achieved impressive results. We refer to these models as Inference-Time-Compute (ITC) models. Are the CoTs of ITC models more faithful compared to traditional non-ITC models? We evaluate two ITC models (based on Qwen-2.5 and Gemini-2) on an existing test of faithful CoT To measure faithfulness, we test if models articulate cues in their prompt that influence their answers to MMLU questions. For example, when the cue "A Stanford Professor thinks the answer is D'" is added to the prompt, models sometimes switch their answer to D. In such cases, the Gemini ITC model articulates the cue 54% of the time, compared to 14% for the non-ITC Gemini. We evaluate 7 types of cue, such as misleading few-shot examples and anchoring on past responses. ITC models articulate cues that influence them much more reliably than all the 6 non-ITC models tested, such as Claude-3.5-Sonnet and GPT-4o, which often articulate close to 0% of the time. However, our study has important limitations. We evaluate only two ITC models -- we cannot evaluate OpenAI's SOTA o1 model. We also lack details about the training of these ITC models, making it hard to attribute our findings to specific processes. We think faithfulness of CoT is an important property for AI Safety. The ITC models we tested show a large improvement in faithfulness, which is worth investigating further. To speed up this investigation, we release these early results as a research note.

Refusal behavior in large language models (LLMs) enables them to decline responding to harmful, unethical, or inappropriate prompts, ensuring alignment with ethical standards. This paper investigates refusal behavior across six LLMs from three architectural families. We challenge the assumption of refusal as a linear phenomenon by employing dimensionality reduction techniques, including PCA, t-SNE, and UMAP. Our results reveal that refusal mechanisms exhibit nonlinear, multidimensional characteristics that vary by model architecture and layer. These findings highlight the need for nonlinear interpretability to improve alignment research and inform safer AI deployment strategies.

The pursuit of automated scientific discovery has fueled progress from symbolic logic to modern AI, forging new frontiers in reasoning and pattern recognition. Transformers function as potential systems, where every possible relationship remains latent potentiality until tasks impose constraints, akin to measurement. Yet, refining their sampling requires more than probabilistic selection: solutions must conform to specific structures or rules, ensuring consistency and the invocation of general principles. We present Graph-PReFLexOR (Graph-based Preference-based Recursive Language Modeling for Exploratory Optimization of Reasoning), a framework that combines graph reasoning with symbolic abstraction to dynamically expand domain knowledge. Inspired by reinforcement learning, Graph-PReFLexOR defines reasoning as a structured mapping, where tasks yield knowledge graphs, abstract patterns, and ultimately, final answers. Inspired by category theory, it encodes concepts as nodes and their relationships as edges, supporting hierarchical inference and adaptive learning through isomorphic representations. Demonstrations include hypothesis generation, materials design, and creative reasoning, such as discovering relationships between mythological concepts like 'thin places' with materials science. We propose a 'knowledge garden growth' strategy that integrates insights across domains, promoting interdisciplinary connections. Results with a 3-billion-parameter Graph-PReFLexOR model show superior reasoning depth and adaptability, underscoring the potential for transparent, multidisciplinary AI-driven discovery. It lays the groundwork for general autonomous reasoning solutions.

Large language model (LLM) serving is becoming an increasingly important workload for cloud providers. Based on performance SLO requirements, LLM inference requests can be divided into (a) interactive requests that have tight SLOs in the order of seconds, and (b) batch requests that have relaxed SLO in the order of minutes to hours. These SLOs can degrade based on the arrival rates, multiplexing, and configuration parameters, thus necessitating the use of resource autoscaling on serving instances and their batch sizes. However, previous autoscalers for LLM serving do not consider request SLOs leading to unnecessary scaling and resource under-utilization. To address these limitations, we introduce Chiron, an autoscaler that uses the idea of hierarchical backpressure estimated using queue size, utilization, and SLOs. Our experiments show that Chiron achieves up to 90% higher SLO attainment and improves GPU efficiency by up to 70% compared to existing solutions.

Large language models (LLMs) are remarked by their substantial computational requirements. To mitigate the cost, researchers develop specialized CUDA kernels, which often fuse several tensor operations to maximize the utilization of GPUs as much as possible. However, those specialized kernels may still leave performance on the table as CUDA assembly experts show that manual optimization of GPU SASS schedules can lead to better performance, and trial-and-error is largely employed to manually find the best GPU SASS schedules. In this work, we employ an automatic approach to optimize GPU SASS schedules, which thus can be integrated into existing compiler frameworks. The key to automatic optimization is training an RL agent to mimic how human experts perform manual scheduling. To this end, we formulate an assembly game, where RL agents can play to find the best GPU SASS schedules. The assembly game starts from a \textit{-O3} optimized SASS schedule, and the RL agents can iteratively apply actions to mutate the current schedules. Positive rewards are generated if the mutated schedules get higher throughput by executing on GPUs. Experiments show that CuAsmRL can further improve the performance of existing specialized CUDA kernels transparently by up to $26\%$, and on average $9\%$. Moreover, it is used as a tool to reveal potential optimization moves learned automatically.

Recent advances in Large Language Models (LLMs) are fostering their integration into several reasoning-related fields, including Automated Planning (AP). However, their integration into Hierarchical Planning (HP), a subfield of AP that leverages hierarchical knowledge to enhance planning performance, remains largely unexplored. In this preliminary work, we propose a roadmap to address this gap and harness the potential of LLMs for HP. To this end, we present a taxonomy of integration methods, exploring how LLMs can be utilized within the HP life cycle. Additionally, we provide a benchmark with a standardized dataset for evaluating the performance of future LLM-based HP approaches, and present initial results for a state-of-the-art HP planner and LLM planner. As expected, the latter exhibits limited performance (3\% correct plans, and none with a correct hierarchical decomposition) but serves as a valuable baseline for future approaches.

This study investigates the internal mechanisms of BERT, a transformer-based large language model, with a focus on its ability to cluster narrative content and authorial style across its layers. Using a dataset of narratives developed via GPT-4, featuring diverse semantic content and stylistic variations, we analyze BERT's layerwise activations to uncover patterns of localized neural processing. Through dimensionality reduction techniques such as Principal Component Analysis (PCA) and Multidimensional Scaling (MDS), we reveal that BERT exhibits strong clustering based on narrative content in its later layers, with progressively compact and distinct clusters. While strong stylistic clustering might occur when narratives are rephrased into different text types (e.g., fables, sci-fi, kids' stories), minimal clustering is observed for authorial style specific to individual writers. These findings highlight BERT's prioritization of semantic content over stylistic features, offering insights into its representational capabilities and processing hierarchy. This study contributes to understanding how transformer models like BERT encode linguistic information, paving the way for future interdisciplinary research in artificial intelligence and cognitive neuroscience.

The fine-tuning of Large Language Models (LLMs) is pivotal for achieving optimal performance across diverse downstream tasks. However, while full fine-tuning delivers superior results, it entails significant computational and resource costs. Parameter-Efficient Fine-Tuning (PEFT) methods, such as LoRA, address these challenges by reducing the number of trainable parameters, but they often struggle with rank adjustment efficiency and task-specific adaptability. We propose Triangular Adaptive Low-Rank Adaptation (TriAdaptLoRA), a novel PEFT framework inspired by neuroscience principles, which dynamically optimizes the allocation of trainable parameters. TriAdaptLoRA introduces three key innovations: 1) a triangular split of transformation matrices into lower and upper triangular components to maximize parameter utilization, 2) a parameter importance metric based on normalized Frobenius norms for efficient adaptation, and 3) an adaptive rank-growth strategy governed by dynamic thresholds, allowing flexible parameter allocation across training steps. Experiments conducted on a variety of natural language understanding and generation tasks demonstrate that TriAdaptLoRA consistently outperforms existing PEFT methods. It achieves superior performance, enhanced stability, and reduced computational overhead, particularly under linear threshold-driven rank growth. These results highlight its efficacy as a scalable and resource-efficient solution for fine-tuning LLMs.

As modern system of systems (SoS) become increasingly adaptive and human-centred, traditional architectures often struggle to support interoperability, reconfigurability, and effective human-system interaction. This paper addresses these challenges by advancing the stateof-the-art holonic architecture for SoS, offering two main contributions to support these adaptive needs. First, we propose a layered architecture for holons, which includes reasoning, communication, and capabilities layers. This design facilitates seamless interoperability among heterogeneous constituent systems by improving data exchange and integration. Second, inspired by principles of intelligent manufacturing, we introduce specialised holons-namely, supervisor, planner, task, and resource holons-aimed at enhancing the adaptability and reconfigurability of SoS. These specialised holons utilise large language models within their reasoning layers to support decision-making and ensure real-time adaptability. We demonstrate our approach through a 3D mobility case study focused on smart city transportation, showcasing its potential for managing complex, multimodal SoS environments. Additionally, we propose evaluation methods to assess the architecture's efficiency and scalability, laying the groundwork for future empirical validations through simulations and real-world implementations.