Research map / Vision-language and multimodal / Multimodal LLMs
C3-Bench: A Context-Aware Change Captioning Benchmark
JAEWOO KIM, Hyeongbeom Kim, Ue-Hwan Kim
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 change criteria to the prompt (ablation) | GPT-5.2 · C3-Bench | Aggregation: 4.93 -> 5.51 | Improved |
| add sequential image tags (ablation) | GPT-5.2 · C3-Bench | Aggregation: 5.35 -> 5.51 | Improved |
| present before-image first (forward) instead of after-image first (ablation) | GPT-5.2 · C3-Bench | Aggregation: 5.31 -> 5.51 | Improved |
| include Change to Detect in the criteria (ablation) | GPT-5.2 · C3-Bench | Aggregation: 5.12 -> 5.51 | Improved |
| include Change to Ignore in the criteria (ablation) | GPT-5.2 · C3-Bench | Aggregation: 5.44 -> 5.51 | Improved |
| include Tone and Nuance in the criteria (ablation) | GPT-5.2 · C3-Bench | Fluency: 8.17 -> 8.65 | Improved |
| add sequential tags and criteria (ablation) | GPT-5.2 · C3-Bench | Reversibility: 0.58 -> 0.62 | Improved |
| train conventional model on all datasets instead of LEVIR-CC only | CLIP4IDC · C3-Bench | Aggregation: 1.79 -> 1.31 | Hurt |
| scale open-source LMM from 3B to 7B | Qwen2.5-VL · C3-Bench | Aggregation: 3.00 -> 4.41 | Improved |
| scale InternVL3 from 2B to 78B | InternVL3 · C3-Bench | Aggregation: 3.35 -> 4.94 | Improved |
| scale Qwen2.5-VL from 32B to 72B | Qwen2.5-VL · C3-Bench | Reversibility: 0.40 -> 0.38 | No clear effect |
| switch from conventional model to criteria-prompted LMM | VARD vs GPT-5.2 · C3-Bench | Aggregation: 2.40 -> 5.51 | Improved |
Closest papers by meaning
- Change3D: Revisiting Change Detection and Captioning from A Video Modeling Perspective (CVPR 2025 · similarity 0.90)
- Region-aware Difference Distilling with Attribute-guided Contrastive Regularization for Change Captioning (AAAI 2025 · similarity 0.89)
- Revisiting Change Captioning from Self-supervised Global-Part Alignment (AAAI 2025 · similarity 0.88)
- VisMin: Visual Minimal-Change Understanding (NeurIPS 2024 · similarity 0.87)
- UniChange: Unifying Change Detection with Multimodal Large Language Model (CVPR 2026 · similarity 0.86)
- CAPability: A Comprehensive Visual Caption Benchmark for Evaluating Both Correctness and Thoroughness (NeurIPS 2025 · similarity 0.86)
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.
