Research map / Vision-language and multimodal / Multimodal LLMs
Mitigating Sycophancy in Multimodal Chart Understanding via Vision-Grounded Verification
Xiaolong Wang, Zhiwei Lin, Tai Liu, Qiang Li, Jian Sun
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 text-only (blind) claim decomposition instead of VLM decomposition (ablation) | Re-Check (Qwen3-VL-8B) · ChartHal overall | accuracy (%): 68.08 -> 72.50 | Improved |
| add CVDS re-calibration (ablation) | Re-Check (Qwen3-VL-8B) · ChartHal overall | accuracy (%): 72.03 -> 72.50 | No clear effect |
| add CVDS re-calibration (ablation) | Re-Check (Qwen3-VL-8B) · ChartHal Normal queries | accuracy (%): 43.51 -> 44.77 | Improved |
| add uncertainty-aware three-way routing instead of two-way (ablation) | Re-Check (Qwen3-VL-8B) · ChartHal overall | accuracy (%): 62.62 -> 72.5 | Improved |
| raise CVDS threshold from 2.0 to 4.0 (ablation) | Re-Check (Qwen3-VL-8B) · ChartHal overall | accuracy (%): 72.5 -> 72.03 | Hurt |
| raise CVDS threshold from 0.0 to 1.0 (ablation) | Re-Check (Qwen3-VL-8B) · ChartHal Normal queries | accuracy (%): 43.51 -> 43.93 | Improved |
| apply Re-Check | Qwen3-VL-8B · ChartHal overall | accuracy (%): 57.49 -> 72.50 | Improved |
| apply Re-Check | InternVL3.5-8B · ChartHal overall | accuracy (%): 13.75 -> 63.47 | Improved |
| apply Re-Check | Qwen3-VL-32B · ChartHal overall | accuracy (%): 53.95 -> 77.78 | Improved |
| use staged Re-Check instead of single-pass integrated prompting | Qwen3-VL-8B · ChartHal Normal queries | accuracy (%): 34.31 -> 44.77 | Improved |
| apply VCD decoding | Qwen3-VL-8B · ChartHal overall | accuracy (%): 57.49 -> 58.29 | No clear effect |
| apply Re-Check | Qwen3-VL-8B · ChartQAPro overall | accuracy (%): 37.37 -> 36.31 | Hurt |
Closest papers by meaning
- ChartNet: A Million-Scale, High-Quality Multimodal Dataset for Robust Chart Understanding (CVPR 2026 · similarity 0.89)
- ChartSketcher: Reasoning with Multimodal Feedback and Reflection for Chart Understanding (NeurIPS 2025 · similarity 0.89)
- CharXiv: Charting Gaps in Realistic Chart Understanding in Multimodal LLMs (NeurIPS 2024 · similarity 0.88)
- ChartMuseum: Testing Visual Reasoning Capabilities of Large Vision-Language Models (NeurIPS 2025 · similarity 0.88)
- Chart-FR1: Visual Focus-Driven Fine-Grained Reasoning on Dense Charts (CVPR 2026 · similarity 0.88)
- EvoChart: A Benchmark and a Self-Training Approach Towards Real-World Chart Understanding (AAAI 2025 · similarity 0.87)
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.
