Research map / Vision-language and multimodal / Multimodal LLMs
Disentangling Pictorial Cue Understanding from Language Bias in VLMs via Depth Ordering Task
Yiqian Liu, Iuliia Kotseruba, John K. Tsotsos
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 few-shot ICL and CoT prompting (ablation) | GPT4.1-mini · O3-D | accuracy: 0.5371 -> 0.6045 | Improved |
| add few-shot ICL and CoT prompting (ablation) | Gemini2.5 · O3-D | accuracy: 0.4728 -> 0.5124 | Improved |
| add few-shot ICL and CoT prompting (ablation) | DeepSeek-VL · O3-D | accuracy: 0.5264 -> 0.5040 | Hurt |
| add few-shot ICL and CoT prompting (ablation) | Qwen2-VL · O3-D | accuracy: 0.5248 -> 0.5128 | Hurt |
| add few-shot ICL and CoT prompting (ablation) | Phi3.5 · O3-D | accuracy: 0.5136 -> 0.5072 | No clear effect |
| add few-shot ICL and CoT prompting (ablation) | VLMs · Relative Size cue | accuracy: 0.5522 -> 0.6214 | Improved |
| add few-shot ICL and CoT prompting (ablation) | VLMs · Familiar Size cue | accuracy: 0.5379 -> 0.5021 | Hurt |
| add linear perspective cue to height-in-plane and light-shadow (ablation) | VLM mean · O3-D HP/LS | accuracy: 0.5444 -> 0.6100 | Improved |
| add linear perspective cue to height-in-plane and focusness (ablation) | VLM mean · O3-D HP/FO | accuracy: 0.5405 -> 0.5193 | Hurt |
| add more depth cues (0-cue to 2-cue images) | VLM mean · O3-D | accuracy: 0.4865 -> 0.5282 | Improved |
| add more depth cues (0-cue to 2-cue images) | DepthAnythingV2 · O3-D | accuracy: 0.5177 -> 0.7446 | Improved |
| increase referring clarity from low to highest | VLM mean · O3-D synthetic | accuracy: 0.4807 -> 0.5068 | Improved |
Closest papers by meaning
- See It from My Perspective: How Language Affects Cultural Bias in Image Understanding (ICLR 2025 · similarity 0.91)
- Can We Talk Models Into Seeing the World Differently? (ICLR 2025 · similarity 0.91)
- On the Robustness of Language Guidance for Low-Level Vision Tasks: Findings from Depth Estimation (CVPR 2024 · similarity 0.90)
- Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective Mitigation (ICML 2026 · similarity 0.90)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.90)
- DepthCues: Evaluating Monocular Depth Perception in Large Vision Models (CVPR 2025 · 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.
