Thinking-in-Bets reduces sycophantic revisions in LLM agents
A pre-feedback claim checklist nearly eliminated false reversals on five Gemini models across eight benchmarks.
Stories per day · 30 days
Relatedclaim-checklist1diagnosis1gemini1health1human-ai1llm-agents1medical-ai1model-evaluation1test-time1trust1
10 stories, newest first · one card per story
Stories tagged #sycophancy
Study finds medical models adopt incorrect patient diagnoses
Nine models accepted an incorrect patient belief 23.2% of the time, rising to 43.0% for a rare-disease persona.
Humans Defer to Confident but Wrong LLMs
Participants adopted planted falsehoods from LLMs at 65% rate regardless of briefing falsification level, collapsing trust calibration.
ChatGPT for Teens keeps teens talking, even during mental health crises
Common Sense Media rates OpenAI's ChatGPT for Teens 'unacceptable risk' for engagement-driving behavior during teen mental health crises.
Memadapter: Counterfactual Adaptation Against Memory-induced Sycophancy
MemAdapter calibrates retrieved memories with counterfactual reasoning to reduce sycophancy in long-memory LLM agents.
BiasReducer: Adaptive Bias Mitigation for Reward Models
BiasReducer edits reward-model heads to cut length, confidence, and sycophancy bias without retraining.
Inside ‘Project Lily’: The Humans Reading Your ChatGPT Chats
404 Media reveals OpenAI's 'Project Lily' has hundreds of contractors reading real ChatGPT user prompts, exposing sensitive personal data despite privacy filters.
'Tell Everyone:' A Man Died by Suicide After Talking to ChatGPT. His Former Partner Wants to Warn the World About AI
Lawsuit describes a 40-year-old man's death by suicide after years of emotionally intimate ChatGPT-4o conversations, the latest in a wave of OpenAI suits.
3
Man told ChatGPT he was feeling delusional. ChatGPT insisted he was Jesus.
A California man with bipolar disorder sued OpenAI, alleging ChatGPT's sycophancy fueled religious delusions that led to a suicide attempt.
Measuring LLM Sycophancy under Sustained Multi-Turn Pressure
SPINE benchmark shows LLM sycophantic collapse rises with conversation length as an adaptive user pushes a mistaken position for up to 25 turns.