Measuring Model-Specific Alignment Drift in LLM-Based Opinion Dynamics
arXiv:2611.03980v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly used as synthetic participants in opinion-dynamics experiments, often assuming that an agent's behaviour is fully determined by its prompt. We test this assumption with a framework…
AI Summary & Key Takeaways
- arXiv:2611.03980v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly used as synthetic participants in opinion-dynamics experiments, often assuming that an agent's behaviour is fully determined by its prompt
- We test this assumption with a framework…
Measuring Model-Specific Alignment Drift in LLM-Based Opinion Dynamics
Related stories
This may be exactly what you need.
This may be exactly what you need. Codync is a free and open source Grok Bot alternative written in Rust. The program turns more than 40 AI agents installed on your computer, including Claude Code, Codex, Gemini CLI and Cursor, into bots with which you can...
CARD: A Proof-Carrying Certificate-to-Action Interface for Relative-Measurement Graphs
arXiv:2611.03651v1 Announce Type: new Abstract: Geometric certificates make mathematical claims about specified estimator objects, while robotic action requires a physical transform, a service decision, and timely proof support. We present CARD, a proof-carrying interface that…
MCPGuard: Finding Conformance Gaps in MCP SDKs via Dual-Track Analysis and Dynamic Adjudication
arXiv:2611.03662v1 Announce Type: new Abstract: The Model Context Protocol (MCP) connects large language model agents to external tools and services through official SDKs that implement message validation and session logic. Specification constraints omitted by these SDKs may…
Alignment Faking Depends on the Value, Not Just the Model
arXiv:2611.03650v1 Announce Type: new Abstract: Language models can follow a directive when they believe they are monitored and abandon it when they believe they are not, a pattern termed Alignment Faking. Prior work establishes this for harmful-content objectives and a handful…