multimodal-model-asked-object-existence-question-over-set-images-onceWhen a multimodal model is asked an object-existence question over a set of images at once, it is less accurate than when the same images are queried one at a time and the answers combined, so cross-image integration adds errors beyond the per-image perception errors.
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: Grounding answers in the image
Observed on
Object-centric questions (existence, counting, attribute, position) over sets of 2-10 natural images; shown for existence in a controlled decomposition ablation across 29 MLLMs inc.
Sources
- Ablation on the existence task comparing comprehensive multi-image questions against per-image decomposed questions over identical content; the paper reports the single-image route wins consistently across all models and scales, but the gap size is only shown in a figure. Overall benchmark average was 36.1% across 29 models; accuracy fell as image count rose (existence 62.4% easy to 30.0% at larger image counts).
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
- For short-answer object-existence questions in open multimodal models (LLaVA v1.5, MiniGPT-4, Qwen2.5-VL), object hallucination is driven partly by the visual encoder itself rather than only by language priors: hallucinated cases show lower image-text embedding similarity and inverted attention, and swapping in a weaker or noisier visual encoder lowers POPE accuracy while a stronger one raises it.Grounding answers in the image · unreviewed
- When a multimodal model is given external text that contradicts the image, making it commit to a description of the image before the text is admitted, then reconciling the two in a second call, recovers accuracy the single-prompt setup loses — but the benefit of withholding the text from the first call is model- and text-source dependent, and on some models context-preserving prompting is still better.Telling the user what they want to hear · unreviewed
- Tool-augmented multimodal models trained only on how to call vision tools accept tool outputs uncritically; adding supervised trajectories where tool evidence is checked against the image and re-queried, plus a reward for correction that reaches a correct answer, raises accuracy on perception-heavy benchmarks (counting, depth, spatial relations) in 3B and 7B Qwen2.5-VL, while slightly lowering it on knowledge-heavy ones.Fixing its own mistakes · unreviewed
- For detecting ungrounded object mentions in open-ended captions from white-box vision-language models, a single internal support signal (token likelihood, attention share, or image-text similarity) confuses support that comes from object-specific image evidence with support that comes from scene plausibility or the generated text prefix; scoring both whether image tokens carry object-specific evidence and whether support is more image-derived than prefix-derived separates these and raises detection AUROC.Grounding answers in the image · unreviewed
- When a video-understanding instruction contains several conditional branches and the model must pick the branch matching what the video shows, both proprietary and open multimodal models pick correctly far more often when the correct branch is listed first, and accuracy falls as the correct branch moves later in the list.Following an unfamiliar procedure · unreviewed
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 model whose accuracy on joint multi-image existence questions matches or exceeds the combined per-image decomposed queries on the same images. Proposed technique, not yet catalogued: Decompose multi-image questions into per- image queries.