Preprint probes how source attribution shifts LLM argument ratings

An unreviewed preprint tested 20 arguments under four source conditions across GPT-5.6 Sol, Claude Opus 5.5, Gemini 4 Argon and DeepSeek Flash. Credibility was most source-sensitive, several models showed an irrelevant-expertise penalty, and agreement shifted less.

• Survey finds leading AI models converge on the same "controversial" take about humans2026-10-22
• User has Claude hand-edit Gemma neurons to tune Italian translation2026-11-03
• StateBench tests long-context state tracking in LLMs2026-11-03
• Claude Opus 5.5 Can Draw Permit-Level Architectural Drawings2026-11-02
• Google October AI recap: Gemini 4 Argon and Genie 4 released2026-11-02