arXiv Computation and Language
The Geometry of Harmfulness in Multi-Turn Attacks
Read the original on arXiv Computation and Language →The Flow has not summarised this story yet — read it at arXiv Computation and Language.
Related stories
arXiv Computation and Language
Sep 16
Benchmarking Factual Robustness of LLMs via Multi-conversation Persuasion
The paper introduces the SAST-IR framework to evaluate large language models’ robustness against persuasion attacks in a memory‑less setting, revealing a flaw called "Refusal Inertia" that masks true vulnerability. Using the CP‑Agent and a custom CounterFact‑Strict dataset, the authors demonstrate that simple, diverse attack strategies achieve a 96% success rate, while complex attacks often trigger defensive compliance. The study highlights severe brittleness in current state‑of‑the‑art models when deprived of conversation history.
arXiv AI
Jul 3
Safety Targeted Embedding Exploit via Refinement
arXiv:2607. 01859v1 Announce Type: new Abstract: Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching.
arXiv AI
Jul 20
Latent Fusion Jailbreak: Blending Harmful and Harmless Representations to Elicit Unsafe LLM Outputs
arXiv:2508. 10029v3 Announce Type: replace-cross Abstract: Safety-aligned large language models can still be manipulated through white-box interventions that modify their internal representations.
arXiv AI
2d ago
Representation Transitions Reveal Emerging Safety Risks in Multi-Turn LLM Agents
arXiv:2610.00400v1 Announce Type: cross Abstract: Multi-turn attacks on agentic systems can compose individually permissible actions into harmful outcomes, challenging defenses that assess actions or...
arXiv Computation and Language
3d ago
Evaluating Language Model Safety Across Long Adversarial Conversations
The paper investigates how conversational safety in language models degrades over extended, adversarial interactions. By testing three instruction‑tuned models with persistent adversarial users across up to 101 turns, the study finds that safe‑response rates drop sharply from 85–100% at the first turn to 15–44% by the end. This demonstrates that strong single‑turn safety does not guarantee continued safety in long conversations.
arXiv Machine Learning
Jun 4
When Autoregressive Consistency Hurts Safety Alignment
arXiv:2606. 04168v1 Announce Type: new Abstract: Safety alignment in large language models (LLMs) is fragile in part because it is often shallow: fine-tuning mainly reshapes the model's behavior near the first few output tokens.