The paper introduces AMRA, a weight‑editing technique that mitigates abliteration—an attack that removes refusal capabilities from large language models by projecting weight matrices orthogonal to a refusal direction. AMRA obscures the refusal signal through rank‑$k$ updates to residual stream writer matrices, replaces refusal‑inducing activations with random aliases, and adjusts downstream reader matrices to maintain original behavior. Experiments on Llama‑3‑8B and Gemma‑2‑9B show significant improvements in post‑abliteration refusal scores with minimal impact on overall model performance.
arXiv:2508.20766v2 Announce Type: replace-cross Abstract: Safety alignment in Large Language Models (LLMs) often involves mediating internal representations to refuse harmful requests. Recent researc...
The paper introduces Latent Space Refusal Anchoring (LSR‑Anchoring), a training‑free technique that extracts a refusal direction from English prompts and applies it to the residual stream of instruction‑tuned models at inference time. The primary variant, Mean‑Activation Steering (MAS), works across several architectures (Llama‑3‑8B, Llama‑3.1‑70B, Mistral‑7B‑Instruct, Qwen2.5‑7B), restoring safety for low‑resource African languages with minimal performance loss, while a refined SAE‑Derived Steering (SDS) further reduces KL divergence without degrading legitimate prompt performance. The method shows positive transfer for Yoruba, Igbo, Igala, and Hausa, but fails for Arabic, suggesting a geometric mismatch rather than a data scarcity issue.
arXiv:2511. 11041v2 Announce Type: replace-cross Abstract: We find that current sentence-embedding models produce outputs with a consistent bias: every embedding $e$ decomposes as $\tilde e + \mu$, where the mean $\mu$ is near-identical across all sentences.
arXiv:2607. 29539v1 Announce Type: cross Abstract: Standard AI-text detection benchmarks compare human-written text against text generated directly by large language models (LLMs).
arXiv:2606. 01196v1 Announce Type: cross Abstract: Safety alignment learned in high-resource languages transfers poorly to low-resource languages.
arXiv:2607. 14111v1 Announce Type: cross Abstract: Can small language models detect and report on perturbations their own internal activations?
The study investigates whether open‑weight language models can introspect on their own internal computations. Using the Open‑Weight Masked Introspection (OWMI) framework, researchers intervened on various internal components of eight models and asked them to report whether changes had occurred. Across 78,000 measurements, none of the models reliably distinguished real interventions from sham ones, with AUROC values essentially at chance. Why It Matters: The findings suggest that current open‑weight models lack the ability to audit their own internal states, highlighting a limitation for oversight that relies on a model’s self‑reporting.
CopyShield is a benchmark that compares three copyright‑defense methods—contrastive decoding, Direct Preference Optimization (DPO), and activation intervention—across two large language models (LLaMA‑3.1‑8B and Mistral‑7B‑v0.3). The study uses controlled memorization of five public‑domain books to measure literal leakage, calibrated non‑literal leakage, utility, and degeneracy, finding that each intervention level yields distinct compliance‑utility trade‑offs. Results show contrastive decoding limits degeneracy but hits a suppression floor, DPO nearly eliminates literal leakage yet causes paraphrase‑loop degeneracy, and activation intervention blocks most non‑literal queries before generation, with human evaluation highlighting coherence and perceived copyright risk differences.
arXiv:2606. 31591v1 Announce Type: cross Abstract: Emergent misalignment (EM) is a recently discovered phenomenon in LLMs where fine-tuning on a narrow misaligned task, such as writing insecure code, leads to broadly misaligned behaviour on unrelated prompts.
arXiv:2507. 02778v3 Announce Type: replace-cross Abstract: Although large language models (LLMs) have transformed AI, they still make errors and follow unproductive reasoning paths.
arXiv:2605. 21706v2 Announce Type: replace Abstract: Safety-aligned language models are trained to refuse harmful requests, yet refusal behavior can be suppressed by steering their internal representations.
The paper investigates how refusal training shapes the internal geometry of language models, showing that activation updates from refusal-completion losses create a distinct low‑dimensional refusal subspace. In a case study on OLMo‑2‑0425‑1B‑Instruct, the authors link the brittleness of refusal directions to repetitive refusal prefixes and demonstrate that using diverse refusal starts can increase the stable rank of gradients, thereby hardening the model against vector‑ablation attacks. The work provides insights into the emergence of safety‑critical features and offers a potential strategy to strengthen refusal robustness.