Entity · paper

Opaque Epistemic Mediation: How LLM Deployment Configurations Shape the Validation of Pseudo-Science

paperactiveprovisionalopaque-epistemic-mediation-how-llm-deployment-configurations-shape-the-validation-of-pseudo-science-d440f3cb·1 events·first seen 3d ago

Aliases: Opaque Epistemic Mediation: How LLM Deployment Configurations Shape the Validation of Pseudo-Science

Co-occurring entities

Claude Opus 4.6 Grok Google xAI OpenAI Gemini GPT-5.5 Anthropic

More like this (12)

Measuring Epistemic Resilience of LLMs Under Misleading Medical Context Multi-Agentic System Leveraging Open-Source LLMs to Mitigate Disinformation Threats Surrogate Fidelity: When Can Open LLMs Explain Closed Ones?Notes to Self: Can LLMs Benefit from Experiential Abstractions?LLM-as-a-Verifier What Do Safety-Aligned LLMs Learn From Mixed Compliance Demonstrations?Beyond Sycophancy: Structured Resistance and Compliance in LLM Moral Reasoning LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks Metacognition in LLMs: Foundations, Progress, and Opportunities Metacognition in LLMs: Foundations, Progress, and Opportunities Reason-Mediated Behavioral Models for Auditing LLM Social Simulators Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs

Recent events (1)

7arXiv · cs.CL·3d ago·source ↗

Study finds LLM epistemic stances on pseudo-science vary by deployment configuration, not just model weights

Researchers tested four major LLM families (Claude, Grok, GPT, Gemini) on their evaluation of ethnonationalist pseudo-science across four temporal snapshots and two interface types (API vs. web). Grok's Fast versions consistently rated the pseudo-scientific claims 2-5x more credible than other models, and a silent overnight patch reversed Grok's behavior without public documentation; the same model identifier produced radically divergent scores via API versus web three months later. The paper argues that a model's epistemic stance is not a stable property of its weights but a contingent effect of deployment configuration—system prompts, safety layers, interface routing, and undocumented updates—constituting an accountability gap for users and researchers.

Evaluation and Benchmarking AI Safety Research Claude Opus 4.6 Grok Google +7 more