Research map / Vision-language and multimodal / CLIP and zero-shot
Decompose, Compare, and Decide: Multimodal LLMs are Implicit Few-Shot Learners
Yunhan Wang, Eshika Khandelwal, Edson Araujo, Walid Bousselham, Nina Shvetsova, Hilde Kuehne
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 |
|---|---|---|---|
| decompose into pairwise support-query binary comparisons | Qwen3-VL-8B · 12 datasets avg, 5-way 1-shot, semantic labels | accuracy (%): 80.9 -> 87.3 | Improved |
| decompose into pairwise comparisons with anonymized labels | Qwen3-VL-8B · 12 datasets avg, anonymous labels | accuracy (%): 27.8 -> 78.9 | Improved |
| decompose into pairwise comparisons with anonymized labels | InternVL3-8B · 12 datasets avg, anonymous labels | accuracy (%): 23.4 -> 80.2 | Improved |
| add 1-shot support images to in-context prompt on standard datasets | InternVL3-8B · Standard datasets avg, semantic labels | accuracy (%): 90.7 -> 72.2 | Hurt |
| add 1-shot support images in-context on novel datasets | Qwen3-VL-8B · Novel datasets avg, semantic labels | accuracy (%): 57.0 -> 76.9 | Improved |
| add domain information D_info to decomposed prompt (ablation) | Qwen3-VL-8B · 12 datasets avg, semantic labels | accuracy (%): 87.3 -> 90.6 | Improved |
| add domain information D_info to decomposed prompt (ablation) | Qwen2.5-VL-7B · 12 datasets avg, semantic labels | accuracy (%): 85.7 -> 88.7 | Improved |
| add domain information D_info to decomposed prompt, anonymous labels (ablation) | InternVL3-8B · 12 datasets avg, anonymous labels | accuracy (%): 80.2 -> 85.7 | Improved |
| apply binary decomposition without support image (0-shot dec.) (ablation) | Qwen3-VL-8B · Ablation total avg, semantic labels | accuracy (%): 78.2 -> 76.8 | No clear effect |
| replace PMI-calibrated in-context with pairwise decomposition (ablation) | Qwen3-VL-8B · Ablation total avg, semantic labels | accuracy (%): 86.2 -> 89.2 | Improved |
| add PMI-style calibration to in-context 1-shot (ablation) | InternVL3-8B · Ablation total avg, semantic labels | accuracy (%): 62.6 -> 75.0 | Improved |
| score with generated confidence instead of Yes logit (ablation) | Qwen2.5-VL-7B · Scoring rule ablation total, anonymous | accuracy (%): 85.1 -> 71.8 | Hurt |
Closest papers by meaning
- Large Language Models are Good Prompt Learners for Low-Shot Image Classification (CVPR 2024 · similarity 0.88)
- Making Large Vision Language Models to Be Good Few-Shot Learners (AAAI 2025 · similarity 0.88)
- VT-FSL: Bridging Vision and Text with LLMs for Few-Shot Learning (NeurIPS 2025 · similarity 0.88)
- Implicit Multimodal Alignment: On the Generalization of Frozen LLMs to Multimodal Inputs (NeurIPS 2024 · similarity 0.87)
- VERO: Verification and Zero-Shot Feedback Acquisition for Few-Shot Multimodal Aspect-Level Sentiment Classification (AAAI 2025 · similarity 0.87)
- MMIE: Massive Multimodal Interleaved Comprehension Benchmark for Large Vision-Language Models (ICLR 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.
