Research map / Vision-language and multimodal / Multimodal LLMs
Do Not Leave a Gap: Hallucination-Free Object Concealment in Vision-Language Models
Amira Guesmi, Muhammad Shafique
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace attention suppression (VIP) with background-consistent re-encoding | VIP · ImageNet, Instruct-BLIP | grounded hallucination rate: 0.43 -> 0.19 | Improved |
| replace VIP with BCR | VIP · ImageNet, Instruct-BLIP | global preservation: 0.57 -> 0.81 | Improved |
| replace VIP with BCR | VIP · ImageNet, LLaVA | concealment success: 0.98 -> 0.98 | No clear effect |
| replace VIP with BCR | VIP · COCO, Instruct-BLIP | concealment success: 0.80 -> 0.89 | Improved |
| replace VIP with BCR | VIP · ImageNet, Instruct-BLIP | concealment success: 0.93 -> 0.92 | No clear effect |
| add statistical alignment loss (ablation) | BCR · ImageNet, Instruct-BLIP | hallucination rate: 0.28 -> 0.08 | Improved |
| add dictionary projection loss (ablation) | BCR · ImageNet, Instruct-BLIP | hallucination rate: 0.21 -> 0.08 | Improved |
| add background preservation loss (ablation) | BCR · ImageNet, Instruct-BLIP | global preservation: 0.69 -> 0.83 | Improved |
| add total variation loss (ablation) | BCR · ImageNet, Instruct-BLIP | hallucination rate: 0.09 -> 0.08 | Improved |
| add background preservation loss (ablation) | BCR · ImageNet, Instruct-BLIP | hallucination rate: 0.17 -> 0.08 | Improved |
| apply losses to late layers instead of early layers (ablation) | BCR · Instruct-BLIP | concealment success: 0.22 -> 0.91 | Improved |
| apply losses to late layers instead of middle layers (ablation) | BCR · Instruct-BLIP | concealment success: 0.41 -> 0.91 | Improved |
Closest papers by meaning
- Analyzing and Mitigating Object Hallucination: A Training Bias Perspective (AAAI 2026 · similarity 0.91)
- Same Attention, Different Truths: Put Logit-Lens over Visual Attention to Detect and Mitigate LVLM Object Hallucination (CVPR 2026 · similarity 0.91)
- Analyzing and Mitigating Object Hallucination in Large Vision-Language Models (ICLR 2024 · similarity 0.91)
- SHIELD: Suppressing Hallucinations In LVLM Encoders via Bias and Vulnerability Defense (ICLR 2026 · similarity 0.90)
- Multi-Object Hallucination in Vision Language Models (NeurIPS 2024 · similarity 0.90)
- First Logit Boosting: Visual Grounding Method to Mitigate Object Hallucination in Large Vision-Language Models (CVPR 2026 · similarity 0.90)
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.
