black-box-hallucination-detection-using-only-sampled-responses-token-log-probabilities CLM-0026For black-box hallucination detection using only sampled responses and token log-probabilities, semantic-entropy signals go to zero whenever the samples collapse into one semantic cluster — which happens for a large share of hallucinated queries, up to nearly all of them on tasks with confusable entities and figures — while token-level uncertainty still separates many of those cases, so the two signals fail on different queries and combining them detects more than either alone.
Capability: Stating false facts confidently
Observed on
Black-box API access exposing text and token log-probabilities, no trusted reference document; measured on seven QA/extraction/summary benchmarks with four models. Supervised combi.
Sources
- Measured: single-cluster rate among hallucinated queries ranged 39% (AmbigQA) to 99% (Financial Summaries); median TopK uncertainty higher for hallucinated single-cluster queries on six of seven datasets, though 21-56% of them fell below the non-hallucinated median. Stacked led or tied in 11 of 26 model-dataset comparisons and was within 0.05 AUROC of the best in 20 of 26. Authors note the leading method's confidence interval overlaps a competitor on every dataset, so rankings among methods are not conclusive; the complementarity analysis is stronger than the ranking.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- In frozen large vision-language models, token-level hallucination evidence is distributed across the relational, spatial and sequential structure of the model's internal trace rather than contained in any single token's hidden state, so a detector that reads the attention graph over image patches, query and response tokens with interleaved GNN/ViT/GRU modules localizes hallucinated spans better than flat probes on the same activations.Grounding answers in the image · reviewed by AI
- For closed-book hallucination detection, training a small supervised classifier over many heterogeneous uncertainty scores beats the best single score, but the gain comes from the diversity of sampling-based black-box signals: ensembles over only single-generation token-probability features add little, while black-box-only ensembles do nearly as well as the full set.Stating false facts confidently · reviewed by AI
- For 7-13B multimodal models, verifying each object claim with a frozen open-vocabulary detector lowers object hallucination only if evidence is pooled across semantically equivalent phrasings and checked for spatial agreement (the preferred boxes of different queries overlapping on one instance); a single-query maximum score is a weaker signal, and shuffling the spatial evidence across samples drives its discrimination to chance.Grounding answers in the image · reviewed by AI
- For training-free inference-time hallucination mitigation in 7B vision- language models, reported gains on object-hallucination benchmarks are largely inseparable from reduced informativeness: hallucination rate and object recall/coverage move together, so a lower score can mean the model mentioned fewer visual entities rather than grounded better.Whether the measurement made the finding · reviewed by AI
- A ~2M-parameter classifier head that reads hidden states already computed during generation (tapping shallow, middle and deep layers of the generating model) detects response-level hallucination at least as well as specialized hidden-state hallucination detectors, and its detection quality rises with the scale of the base model it is attached to.Stating false facts confidently · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: Show that on single-semantic-cluster hallucinated queries token-uncertainty features give no separation above chance, or that a combined detector never beats the better single signal across datasets and models. Proposed technique, not catalogued: Stacked semantic-plus-token hallucination detector.