AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Mitigating Hallucination in Vision-Language Model with Depth and Spatial-aware Key-Value Refinement

Gusang Lee, Soohyun Kim, Donghoon Kim, Kyuhong Shim, Byonghyo Shim

ICLR 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Decoding-time hallucination mitigation

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Studies the representational origins of visual hallucination in vision-language models and proposes a depth- and spatial-aware key-value refinement to mitigate it.

Problem

Large vision-language models produce content not grounded in the input image (visual hallucination), and the representational origins of these hallucinations remain unclear.

Approach

Analyzes internal representations to show that successful grounding depends on depth and spatial information, and refines key-value representations using depth- and spatial-awareness to reduce hallucination. Full method details are unavailable beyond the abstract.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply DSCR depth and spatial KV cache refinementLLaVA-1.5 · MME hallucination subsettotal score: 892.68 -> 925.96Improved
apply DSCRQwen-VL · MME hallucination subsettotal score: 825.01 -> 872.61Improved
apply DSCRQwen2.5-VL · MME hallucination subsettotal score: 1042.91 -> 1039.97No clear effect
apply DSCRLLaVA-1.5 · CHAIRCHAIR_S: 48.2 -> 39.2Improved
apply DSCRLLaVA-1.5 · CHAIRRecall: 78.2 -> 73.4Hurt
apply DSCRLLaVA-1.5 · RePOPE randomF1: 0.72 -> 0.80Improved
apply DSCRmPLUG-Owl2 · POPE (GQA) randomaccuracy: 0.85 -> 0.84Hurt
apply DSCRLLaVA-1.5 · Depth Hallucination Mini-Benchmarkaverage score: 55.5 -> 78.5Improved
add DSCR on top of OPERAOPERA · Depth Hallucination Mini-Benchmarkaverage score: 75.5 -> 76.0No clear effect
add DSCR on top of VCDVCD · Depth Hallucination Mini-Benchmarkaverage score: 70.5 -> 72.0Improved
combine depth and spatial cues rather than one alone (ablation)DSCR · MME hallucination subsettotal score: 645.00 -> 650.00Improved
refine key cache in addition to value cache (ablation)DSCR · MME hallucination subsettotal score: 640.00 -> 650.00Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy