training-free-inference-time-hallucination-mitigation-7b-vision-language-models-reported-gains CLM-0176For 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.
Capability: Whether the measurement made the finding
Observed on
Six decoding-time and attention/hidden-state mitigation methods on three 7B open LVLMs (LLaVA-1.5, LLaVA-NeXT, InstructBLIP), evaluated on CHAIR, AMBER, and MMStar with author-defa.
Sources
- supportsDoes Playing it Safe Count as Faithfulness? Reassessing LVLM Hallucination Mitigation Methods54 configurations (3 models x 6 methods x 3 benchmarks). Reports Pearson r=0.73 between CHAIRs and object recall, r=0.70 between AMBER Hal and Cover, over the method set. Correlation is across methods, not within a method under varied strength, so it does not isolate a causal mechanism; the authors say so. Two methods (CAAC, CEI) reportedly preserved recall, so the pattern is a tendency, not universal. On MMStar, 16 of 18 fine-grained-perception configurations degraded or gained under 1%.
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
- For training-free inference-time hallucination interventions in 7B-class vision-language models, increasing intervention strength (finer regional evidence partitioning, longer attention-routing windows, stronger textual perturbation) keeps lowering CHAIR object-hallucination scores while producing empty or reduced-coverage responses, so lower hallucination numbers at high intervention strength partly reflect the model saying less rather than grounding better.Grounding answers in the image · reviewed by AI
- 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 · reviewed by AI
- 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 · 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
- In 7B-class LVLMs (LLaVA-1.5, MiniGPT-4, mPLUG-Owl2), hallucination directions in activation space are not a single mode: clustering hallucinated-minus-faithful hidden-state differences into several low-rank subspaces and weighting them per test image lowers object-hallucination rates slightly below a single global null-space projection, with the margin small (CHAIR_I 5.30 to 4.92 on LLaVA-1.5) and inconsistent on yes/no existence probing.Grounding answers in the image · reviewed by AI
Notes
Ingested unreviewed on 2026-09-07 and deliberately inert until a human endorses it: it does not move a technique standing, does not count toward the backtest, and is excluded anywhere a claim would carry weight. Drafted confidence: medium. Falsifier as drafted: A mitigation method that lowers CHAIR/AMBER hallucination while holding or raising object recall and coverage, and does not degrade MMStar fine-grained perception and reasoning, across several models. Automatic check flagged: figures not in the source: 16. Proposed technique, not yet catalogued: Score hallucination jointly with informativeness and general capability.