{"arxiv_id":"2603.19574","incoming":{"papers":5,"sentences":6,"groups":[{"kind":"supports_this_paper","label":"Cites it as support","papers":2,"sentences":2,"quotes":[{"arxiv_id":"2606.00975","title":"Lost in Delusion: Examining LLM Safety Under User Delusions and Distress","url":"/paper/2606.00975","section":"LLMs, Mental Health, and Safety","sentence":"Empirically, delusional prompts are far more likely to elicit inappropriate responses [ref] , sycophancy saturates real human-chatbot logs involving delusions [ref] , and longitudinal delusion-related language increases in simulated extended conversations with frontier models [ref] .","intent_p":0.99,"contests_p":null},{"arxiv_id":"2607.07003","title":"Dissociating the Internal Representations of Sycophancy in LLMs","url":"/paper/2607.07003","section":"Introduction","sentence":"While seeming innocuous, sycophancy can spread misinformation, perpetuate ungrounded user biases, and even contribute to delusional spirals [ref] .","intent_p":0.99,"contests_p":null}]}],"unlabelled_sentences":4},"outgoing":{"papers":12,"sentences":13,"groups":[{"kind":"method_or_tool","label":"Uses the method or tool of","papers":1,"sentences":1,"quotes":[{"arxiv_id":"2203.05794","title":"BERTopic: Neural topic modeling with a class-based TF-IDF procedure","url":"/paper/2203.05794","section":"Themes in Delusion-related Language","sentence":"In particular, we conduct topic modeling using BERTopic [ref] which clusters documents in embedding space and derives topic representations through class-based term weighting.","intent_p":1.0,"contests_p":null}]},{"kind":"background","label":"Cites as background","papers":5,"sentences":5,"quotes":[{"arxiv_id":"2504.18932","title":"AI Chatbots for Mental Health: Values and Harms from Lived Experiences of Depression","url":"/paper/2504.18932","section":"Related Work","sentence":"A growing body of work has studied potential harms in conversational AI systems, including sycophancy and reinforcement of user beliefs [ref] .","intent_p":0.91,"contests_p":null},{"arxiv_id":"2412.00207","title":"Can LLM \"Self-report\"?: Evaluating the Validity of Self-report Scales in Measuring Personality Design in LLM-based Chatbots","url":"/paper/2412.00207","section":"Related Work","sentence":"Another line of work studies how LLMs can mirror user behavior and simulate conversational agents with personalized styles [ref] .","intent_p":0.95,"contests_p":null},{"arxiv_id":"2601.12754","title":"PAIR-SAFE: A Paired-Agent Approach for Runtime Auditing and Refining AI-Mediated Mental Health Support","url":"/paper/2601.12754","section":"Practical and Design Implications","sentence":"Much of the current safety research on conversational AI focuses on mitigating harms in single-turn or single-session interactions, such as unsafe outputs to isolated prompts [ref] .","intent_p":0.92,"contests_p":null},{"arxiv_id":"2601.13235","title":"RubRIX: Rubric-Driven Risk Mitigation in Caregiver-AI Interactions","url":"/paper/2601.13235","section":"Practical and Design Implications","sentence":"Much of the current safety research on conversational AI focuses on mitigating harms in single-turn or single-session interactions, such as unsafe outputs to isolated prompts [ref] .","intent_p":0.92,"contests_p":null},{"arxiv_id":"2504.13277","title":"Interpersonal Theory of Suicide as a Lens to Examine Suicidal Ideation in Online Spaces","url":"/paper/2504.13277","section":"Related Work","sentence":"Online spaces enable candid mental health self-disclosure with communities for peer support and a sense of belonging [ref] .","intent_p":0.81,"contests_p":null}]}],"unlabelled_sentences":7},"claims":[],"thresholds":{"intent_p":0.8,"supports_p":0.95,"contests_p":0.7},"method":"Citing sentences come from each paper's LaTeX source; every \\cite marks the sentence and the works it cites. Pith's classifier typed each sentence. A label is shown at intent probability 0.8 or more (0.95 for support), a contest at 0.7 or more. Measured accuracy against two blind model families is at https://pith.science/accuracy. The quoted sentences are the citing authors' own.","held_up":null,"_links":{"html":"https://pith.science/paper/2603.19574"}}