Search papers, labs, and topics across Lattice.
This paper introduces the Cognitive Circuit Breaker, a framework for detecting hallucinations in LLMs by monitoring the discrepancy between softmax probabilities and internal latent certainty during the forward pass. By calculating the "Cognitive Dissonance Delta" using linear probes on hidden states, the method identifies instances where the model's outward confidence doesn't align with its internal representation. Experiments demonstrate statistically significant hallucination detection with minimal computational overhead, offering an architecture-dependent OOD generalization capability.
LLMs often confidently spout falsehoods, but this "Cognitive Circuit Breaker" can detect those lies by listening to the model's hidden thoughts during inference.
As Large Language Models (LLMs) are increasingly deployed in mission-critical software systems, detecting hallucinations and ``faked truthfulness''has become a paramount engineering challenge. Current reliability architectures rely heavily on post-generation, black-box mechanisms, such as Retrieval-Augmented Generation (RAG) cross-checking or LLM-as-a-judge evaluators. These extrinsic methods introduce unacceptable latency, high computational overhead, and reliance on secondary external API calls, frequently violating standard software engineering Service Level Agreements (SLAs). In this paper, we propose the Cognitive Circuit Breaker, a novel systems engineering framework that provides intrinsic reliability monitoring with minimal latency overhead. By extracting hidden states during a model's forward pass, we calculate the ``Cognitive Dissonance Delta''-- the mathematical gap between an LLM's outward semantic confidence (softmax probabilities) and its internal latent certainty (derived via linear probes). We demonstrate statistically significant detection of cognitive dissonance, highlight architecture-dependent Out-of-Distribution (OOD) generalization, and show that this framework adds negligible computational overhead to the active inference pipeline.
