Research map / Vision-language and multimodal / Human and embodied modeling
Disentangling Hallucinations: Orthogonal Semantic Projection for Robust Interpretability
Emirhan Bilgiç, Baptiste Caramiaux, Zhi Yan, Gianni Franchi
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 |
|---|---|---|---|
| add OSP | AttentionCAM on CLIP · ImageNet-Segmentation | AUROC: 52.68 -> 67.73 | Improved |
| add OSP | CheferCAM on SigLIP · ImageNet-Segmentation | AUROC: 55.47 -> 62.64 | Improved |
| add OSP | GradCAM on CLIP · ImageNet-Segmentation | AUROC: 65.39 -> 71.82 | Improved |
| add OSP | LeGrad on SigLIP · ImageNet-Segmentation | AUROC: 74.42 -> 74.73 | Improved |
| add OSP | DAAM on Stable Diffusion 2 · ImageNet-Segmentation | AUROC: 83.07 -> 85.48 | Improved |
| add OSP | AttentionCAM on CLIP · ImageNet-Segmentation positive prompts | mIoU: 40.14 -> 47.96 | Improved |
| add OSP | LeGrad on CLIP · ImageNet-Segmentation negative prompts | mIoU: 40.88 -> 40.74 | No clear effect |
| add OSP | CheferCAM on SigLIP · ImageNet-Segmentation negative prompts | mIoU: 36.65 -> 38.45 | Hurt |
| add OSP | LeGrad on SigLIP · ImageNet-Segmentation negative prompts | mIoU: 37.27 -> 34.68 | Improved |
| use ImageNet-class dictionary instead of no OSP (ablation) | OSP, avg 9 pairs · ImageNet-Segmentation | AUROC: 70.64 -> 74.08 | Improved |
| use WordNet-augmented dictionary (ablation) | OSP, avg 9 pairs · ImageNet-Segmentation | AUROC: 70.43 -> 73.90 | Improved |
| use Gemini 3 Flash dictionary (ablation) | OSP, avg 9 pairs · ImageNet-Segmentation | AUROC: 70.65 -> 75.52 | Improved |
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.
- GradCAM — Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization (ICCV 2017) · better on 3 of 3 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
- When Semantics Mislead Vision: Mitigating Large Multimodal Models Hallucinations in Scene Text Spotting and Understanding (NeurIPS 2025 · similarity 0.91)
- Interpreting and Editing Vision-Language Representations to Mitigate Hallucinations (ICLR 2025 · similarity 0.91)
- Image Token Matters: Mitigating Hallucination in Discrete Tokenizer-based Large Vision-Language Models via Latent Editing (NeurIPS 2025 · similarity 0.90)
- Mitigating Hallucinations in Large Vision-Language Models by Adaptively Constraining Information Flow (AAAI 2025 · similarity 0.90)
- Reducing Hallucinations in Large Vision-Language Models via Latent Space Steering (ICLR 2025 · similarity 0.90)
- Analyzing and Mitigating Object Hallucination: A Training Bias Perspective (AAAI 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.
