Recovering Agentic Sovereignty: Mitigating the Consensus Paradox via Contrastive Epistemic Decoding

T1Covered by 1 source · first reported by arXiv at 24 Sept, 04:00 UTC

arXiv:2609.25570v1 Announce Type: new Abstract: Large language models (LLMs) exhibit a parametric vulnerability to adversarial swarm consensus. To mitigate this sycophancy, we introduce Contrastive Epistemic Decoding (CED), a zero-shot inference intervention. Unlike standard Contrastive Decoding…

aisciencesecurity

Sources

• T1Recovering Agentic Sovereignty: Mitigating the Consensus Paradox via Contrastive Epistemic DecodingarXiv24 Sept, 04:00