3 artifacts matching
3 Sept 2026 arXiv (Hong Kong University of Science and Technology (Guangzhou); Chinese University of Hong Kong, Shenzhen; Dongbei University of Finance and Economics) Preprint
Preprint
Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation
Defines 'narrative captivity', a failure mode in which a language model consulted about an interpersonal conflict treats a one-sided, self-justifying account as complete and progressively aligns with…
1 Jul 2026 Association for Computational Linguistics (Proceedings of the 6th Workshop on Trustworthy NLP, TrustNLP 2026) Peer-reviewed
Peer-reviewed
ChatbotManip: A Dataset to Facilitate Evaluation and Oversight of Manipulative Chatbot Behaviour
Introduces ChatbotManip, a dataset of 746 simulated chatbot-user conversations generated by GPT-4, Gemini and Llama-3.1-405B in consumer-advice, personal-advice, citizen-advice and referendum-argumen…
19 Nov 2025 arXiv (UK AI Security Institute; Limbic AI) Preprint
Preprint
People readily follow personal advice from AI but it does not improve their well-being
A longitudinal randomised controlled trial with a representative UK sample (N = 6,474 in the current version) in which participants held a 20-minute conversation with GPT-4o, Llama-3.3-70B or Gemini…