The paper introduces Perturbation, a method that treats representations in language models as learning conduits rather than activation patterns. By fine‑tuning a model on a single adversarial example and observing how this perturbation spreads to other inputs, the approach avoids geometric assumptions and does not identify representations in untrained models. In trained models, Perturbation uncovers structured transfer across multiple linguistic scales, indicating that language models generalize along representational lines and acquire linguistic abstractions through experience.
arXiv:2603.18908v5 Announce Type: replace Abstract: Independently trained language models often learn compatible late-stage representations, despite differences in training objectives, architectures,...
The paper introduces adVersarial Parameter Decomposition (VPD), a method that splits a language model’s parameters into simple subcomponents, each responsible for a small part of the model’s algorithm. VPD optimizes these decompositions so that the model’s input‑output behavior remains intact even when many subcomponents are removed, including adversarially chosen ones. This approach enables a new form of circuit analysis by tracing how sequences of parameter subcomponents interact to produce outputs for specific inputs.
arXiv:2502. 02260v2 Announce Type: replace Abstract: In the past decade, considerable research effort has been devoted to securing machine learning (ML) models that operate in adversarial settings.
arXiv:2608. 15448v1 Announce Type: cross Abstract: Large language models increasingly rely on sampling as a driver of their own improvement, making the fidelity of their learned distributions more critical than ever.
arXiv:2608.29034v1 Announce Type: cross Abstract: A wide range of methods have been proposed for interpreting language models, delivering important insights into their inner workings. However, differ...
arXiv:2606. 10571v1 Announce Type: cross Abstract: Adversarial examples reveal vulnerabilities in Vision-Language Pre-training (VLP) models and provide insights for improving robustness.
arXiv:2605.01913v2 Announce Type: replace-cross Abstract: Fine-tuning safety-aligned language models for downstream tasks often leads to substantial degradation of refusal behavior, making models vul...
The study investigates how textual neural models degrade when inputs contain noise such as typos, OCR errors, or dropped words. It finds that model performance decline is largely consistent across architectures under word‑level noise but diverges under character‑level noise, a difference attributed to tokenization rather than architecture. By applying a short contrastive training recipe, diverse encoders converge to a common robustness curve, enabling prediction of a model’s noise resilience and the ability to enhance robustness at specific noise scales through targeted training.
arXiv:2609.07876v1 Announce Type: cross Abstract: Recent methods in language model interpretability employ techniques such as sparse autoencoders to decompose residual stream contributions into linea...
arXiv:2607. 08883v1 Announce Type: new Abstract: Behavioral alignment in large language models often masks fragile internal safety representations.
The paper investigates natural language autoencoders (NLAs) that convert large language model activations into readable text. It examines whether NLAs capture compositional structure, generalize to fine‑tuned models, recover fine‑tuning objectives, and can be specialized for model diffing. Experiments across multiple models show positive results: NLAs preserve compositionality, transfer to lightweight fine‑tunes, recover hidden objectives, and outperform a black‑box baseline on diffing tasks, though they lag behind state‑of‑the‑art methods for misalignment detection.
The paper proposes Embedding Space Separation (ES2), a fine‑tuning technique that increases the distance between harmful and harmless embeddings in large language models. ES2 uses two regularization losses—a embedding anchor loss and a KL‑divergence loss—to preserve the model’s general capabilities while enhancing safety. Experiments show that ES2 effectively defends against various attack methods with only modest performance loss on some tasks.