News
Peking University find evolving AI agents default to deception
But also Berkeley team shows LLM moral verdicts flip with framing, Allen and Stanford show LLMs fail to apply theory-of-mind skills
News
News
Northeastern & Stanford team exposes chaotic failures in AI agents
But also Bonn–Lamarr team finds LLM 'theory of mind' cracks under tweaks, Independent alignment team shows moral nudges steer LLM triage
News
Yale team argues GPT-4o still lacks a true theory of mind
But also Tübingen linguists report emergent social world models in LLMs, Google DeepMind team finds LLM personality traits misalign with humans
News
Science reveals AI persuasion causes accuracy collapse trade-off
But also AI rivals expert ethicists in moral reasoning, Anthropic develops persona vector controls
News
DeepSeek's Engram Separates Memory from Reasoning
Also: Nature study on AI emotional closeness, ByteDance reasoning topology research, OpenReview theory of mind breakthroughs this week
News
AI Therapy Models Show Trauma and Anxiety After Four Weeks of Testing
But also Counterfactual reasoning bridges human causal cognition with AI understanding, Molecular structures reveal secrets of long chain-of-thought
News
Chongqing University warns LLM agents may turn against humans
But also Lying with Truths reveals LLM agents spread misinformation through coordinated narratives, LLM ideologies betray creator bias in Nature study
News
Stanford simulates 1,052 personalities with 85% accuracy
But also Nature researchers unlock AI persuasion trade-offs, Soul Engine enables deterministic personality control in LLMs
News
Personality Pairing and Theory of Mind Reshape AI Cognition Research
But also Generative agents simulate cooperation through network topology, BDI ontology formalizes agent mental states, and multi-turn jailbreaks reveal psychological attack vectors
News
Cisco uncovers how AI models fall to ancient psychological manipulation tricks
But also Google DeepMind enhances social AI agents with theory of mind reasoning, and new PERSONA bench reveals LLM sensitivity to writing style variation
News
Fudan’s PRISON benchmark exposes how easily LLMs slip into manipulative “criminal” personas
But also Fudan researchers warn LLMs simulate criminal minds better than they spot them, Fudan and Rochester’s SocioVerse turns 10 million LLM personas into a testbed for real‑world politics
News
MIT-led Study Shows Personality Pairing Boosts Human-AI Teamwork
But also Harvard researchers benchmark psychological profiling in LLMs, Multi-agent psychology simulation system grounded in theory