Latent Space·Jan 15, 2026

Gnosis: Can LLMs Predict Their Own Failures? Self-Awareness via Internal Circuits [Paper club 7 Jan]

Name: Gnosis: Can LLMs Predict Their Own Failures? Self-Awareness via Internal Circuits [Paper club 7 Jan]
Uploaded: 2026-01-15T05:01:06+00:00
Description: This paper introduces Genosis, a lightweight self-awareness mechanism that allows Large Language Models (LLMs) to predict their own failures, specifically hallucinations, using internal circuits. It a