Large Language Models (LLMs) are vulnerable to adversarial attacks that bypass safety guidelines and generate harmful content. Mitigating these vulnerabilities requires defense mechanisms that are both robust and computationally efficient. However, existing approaches either incur high computational costs or rely on lightweight defenses that can be easily circumvented, rendering them impractical for real-world LLM-based systems. In this work, we introduce the AlignTree defense, which enhances model alignment while maintaining minimal computational overhead. AlignTree monitors LLM activations during generation and detects misaligned behavior using an efficient random forest classifier. This classifier operates on two signals: (i) the refusal direction - a linear representation that activates on misaligned prompts, and (ii) an SVM-based signal that captures non-linear features associated with harmful content. Unlike previous methods, AlignTree does not require additional prompts or auxiliary guard models. Through extensive experiments, we demonstrate the efficiency and robustness of AlignTree across multiple LLMs and benchmarks.

Attention matrices are fundamental to transformer research, supporting a broad range of applications including interpretability, visualization, manipulation, and distillation.Yet, most existing analyses focus on individual attention heads or layers, failing to account for the model's global behavior.While prior efforts have extended attention formulations across multiple heads via averaging and matrix multiplications or incorporated components such as normalization and FFNs, a unified and complete representation that encapsulates all transformer blocks is still lacking.We address this gap by introducing TensorLens, a novel formulation that captures the entire transformer as a single, input-dependent linear operator expressed through a high-order attentioninteraction tensor.This tensor jointly encodes attention, FFNs, activations, normalizations, and residual connections, offering a theoretically coherent and expressive linear representation of the model's computation.TensorLens is theoretically grounded and our empirical validation shows that it yields richer representations than previous attention-aggregation methods.Our experiments demonstrate that the attention tensor can serve as a powerful foundation for developing tools aimed at interpretability and model understanding.

Causal masking is a fundamental component in Generative Pre-Trained Transformer (GPT) models, playing a crucial role during training.Although GPTs can process the entire user prompt at once, the causal masking is applied to all input tokens step-by-step, mimicking the generation process.This imposes an unnecessary constraint during the initial "prefill" phase when the model processes the input prompt and generates the internal representations before producing any output tokens.In this work, attention is masked based on the known block structure at the prefill phase, followed by the conventional token-by-token autoregressive process after that.For example, in a typical chat prompt, the system prompt is treated as one block, and the user prompt as the next one.Each of these is treated as a unit for the purpose of masking, such that the first tokens in each block can access the subsequent tokens in a non-causal manner.Then, the model answer is generated in the conventional causal manner.The Segment-by-Segment scheme entails no additional computational overhead.When integrated using a lightweight fine-tuning into already trained models such as Llama and Qwen, MAS quickly increases models' performances.Our code will be available at: https://github.com/shacharKZ/ MAS-Segment-

Shahar Katz, Lior Wolf. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

We present a novel approach to neural code generation that incorporates real-time execution signals into the language model generation process. While large language models (LLMs) have demonstrated impressive code generation capabilities, they typically do not utilize execution feedback during inference, a critical signal that human programmers regularly leverage. Our method, Execution-Guided Classifier-Free Guidance EG-CFG, dynamically incorporates execution signals as the model generates code, providing line-by-line feedback that guides the generation process toward executable solutions. EG-CFG employs a multi-stage process: first, we conduct beam search to sample candidate program completions for each line; second, we extract execution signals by executing these candidates against test cases; and finally, we incorporate these signals into the prompt during generation. By maintaining consistent signals across tokens within the same line and refreshing signals at line boundaries, our approach provides coherent guidance while preserving syntactic structure. Moreover, the method naturally supports native parallelism at the task level in which multiple agents operate in parallel, exploring diverse reasoning paths and collectively generating a broad set of candidate solutions. Our experiments across diverse coding tasks demonstrate that EG-CFG significantly improves code generation performance compared to standard approaches, achieving state-of-the-art results across various levels of complexity, from foundational problems to challenging competitive programming and data science tasks.

Understanding how Transformer-based Language Models (LMs) learn and recall information is a key goal of the deep learning community.Recent interpretability methods project weights and hidden states obtained from the forward pass to the models' vocabularies, helping to uncover how information flows within LMs.In this work, we extend this methodology to LMs' backward pass and gradients.We first prove that a gradient matrix can be cast as a low-rank linear combination of its forward and backward passes' inputs.We then develop methods to project these gradients into vocabulary items and explore the mechanics of how new information is stored in the LMs' neurons.Our code is available