TriLens: Per-Layer Logit-Lens Entropy for White-Box Hallucination Detection
For practitioners needing reliable hallucination detection in instruction-tuned LLMs, TriLens offers a lightweight, interpretable method that leverages internal computation dynamics.
TriLens introduces a white-box hallucination detector for LLMs that tracks entropy of logit-lens readouts from three internal modules per layer, achieving strong detection performance across instruction-tuned models and QA benchmarks without storing high-dimensional states or sampling multiple generations.
When a language model hallucinates, the final answer is wrong, but the mistake is not necessarily invisible inside the model. Different internal pathways may remain uncertain, disagree in how quickly they sharpen, or commit to competing continuations before the output is produced. We introduce TriLens, a white-box detector that turns this intuition into a compact representation: at every layer, it reads the multi-head self-attention output, the feed-forward output, and the residual stream through the model's own logit lens, then records only the entropy of each readout. The resulting 3L-dimensional trajectory describes how certainty forms across depth and across modules, without storing high-dimensional hidden states or sampling multiple generations. This simple signal yields a strong detector across instruction-tuned LLMs and QA benchmarks, and our analyses show that the three module-wise entropy trajectories provide complementary evidence. TriLens suggests that hallucination detection can benefit from tracking how internal computation settles, not only what the final layer predicts.