detecting-hallucinated-text-spans-vision-language-model-outputs-span-localizationFor detecting hallucinated text spans in vision-language model outputs, span localization and confidence calibration come from different architectures: a fine-tuned generative VLM proposes better span boundaries but scores them poorly, while a discriminative token tagger trained with a direct regression loss on human-consensus confidence gives better-correlated probabilities but conservative spans, so using the generative spans as proposals and the tagger's probabilities as scores beats either alone.
Ingested from a paper but not yet reviewed by a human. It is deliberately inert: it does not move any technique’s standing, does not count toward the backtest, and is excluded anywhere a claim would carry weight. Read the source before relying on it.
Capability: Stating false facts confidently
Observed on
Character-level hallucination span detection on the SHROOM-Visions dataset (EN, FR, IT, ZH), responses under 512 tokens; tuning done on the English split..
Sources
- Measured on the English validation split (N=379): tagger Pearson 0.369 vs generative 0.285; generative hallucinated IoU 0.182 vs tagger 0.155; fusion 0.413 Pearson and 0.391 IoU. Ablation shows removing the MSE calibration loss costs the most (-0.095 Pearson). Single task, single team, fusion weights tuned on the same validation split used to report the main number; hallucinated-span IoU stays low (0.196) throughout.
Disagreeing is the most useful thing you can do here. Both sides of every contested claim in this catalog were assembled by the same person, which is its weakest point.
Related claims
- In 7-8B vision-language models, fine-tuning visual and text embeddings toward perturbation-averaged and ground-truth anchors before cross-modal contrastive alignment lowers object and attribute hallucination rates on captioning and existence benchmarks without reducing object coverage or general VQA accuracy.Grounding answers in the image · unreviewed
- 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 · unreviewed
- In diffusion vision-language models generating free-form captions, object tokens that are hallucinated commit at later denoising steps and with lower per-token confidence than grounded object tokens, so the commit-step trajectory is an available internal hallucination signal that has no counterpart in autoregressive decoding.Grounding answers in the image · unreviewed
- In synthetic hallucination benchmarks where faithful items are human-written and hallucinated items are LLM-rewrites, a detector's binary score partly measures human-vs-machine style rather than factuality: feeding faithful LLM rewrites of the same truthful articles makes some fine-tuned encoders flag nearly all of them as hallucinated.Whether the measurement made the finding · unreviewed
- In open-source vision-language models (LLaVA-1.5, Qwen-VL-Chat, Qwen2.5-VL, LLaVA-NeXT), a training-free decoding intervention that detects large layer-to-layer and step-to-step shifts in hidden states and nudges the diverging states back toward their prior value lowers object-hallucination rates, but the lower rate comes with fewer objects mentioned and shorter outputs.Grounding answers in the image · unreviewed
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: low. Falsifier as drafted: A fine-tuned generative VLM whose own token probabilities match or beat a dedicated regression-head tagger on calibration correlation, or a fusion that fails to improve on both subsystems on a held-out split. Proposed technique, not catalogued: union-calibrated fusion of a generative span proposer with a calibrated tagger.