Research map / Vision-language and multimodal / Multimodal LLMs
Otter: A Multi-Modal Model With In-Context Instruction Tuning
Bo Li, Yuanhan Zhang, Liangyu Chen, Jinghao Wang, Fanyi Pu, Joshua Adrian Cahyono, Jingkang Yang, Chunyuan Li, Ziwei Liu
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 answer format to the task description prefix (ablation) | Otter-Image (LLaVA-Instruct ablation) · MME Cognition | score: 1212.9 -> 1414.5 | Improved |
| add question coverage to the task description (ablation) | Otter-Image (LLaVA-Instruct ablation) · MME Cognition | score: 1212.9 -> 1315.2 | Improved |
| add model ability to the task description (ablation) | Otter-Image (LLaVA-Instruct ablation) · MM-Vet | GPT-4 score: 23.7 -> 25.8 | Improved |
| add answer format to the task description (ablation) | Otter-Image (LLaVA-Instruct ablation) · MM-Vet | GPT-4 score: 23.7 -> 22.3 | Hurt |
| add the full task description at evaluation (ablation) | Otter-Image (LLaVA-Instruct ablation) · MM-Vet | GPT-4 score: 23.7 -> 27.0 | Improved |
| train with 2-shot in-context examples (ICIT), evaluate 2-shot (ablation) | Otter-Image · MME Cognition | score: 1310.7 -> 1430.3 | Improved |
| train with 4-shot in-context examples (ICIT), evaluate 4-shot (ablation) | Otter-Image · MME Cognition | score: 1173.5 -> 1440.8 | Improved |
| increase evaluation shots from 2 to 4 without in-context instruction tuning (ablation) | Otter-Image trained w/o ICIT · MME Cognition | score: 1310.7 -> 1173.5 | Hurt |
| add in-context instruction tuning for few-shot COCO Caption (ablation) | Otter-Image · COCO Caption few-shot | captioning score | Improved |
| add LLAVAR data to MIMIC-IT subsets (ablation) | Otter-Image · MMBench | accuracy: 41.5 -> 43.2 | Improved |
| add academic datasets to MIMIC-IT subsets (ablation) | Otter-Image · MMBench | accuracy: 41.5 -> 62.1 | Improved |
| add academic datasets to MIMIC-IT subsets (ablation) | Otter-Image · MM-Vet | GPT-4 score: 19.7 -> 30.6 | Improved |
Closest papers by meaning
- INST-IT: Boosting Instance Understanding via Explicit Visual Prompt Instruction Tuning (NeurIPS 2025 · similarity 0.90)
- Learning to Instruct for Visual Instruction Tuning (NeurIPS 2025 · similarity 0.88)
- MoDA: Modulation Adapter for Fine-Grained Visual Understanding in Instructional MLLMs (ICML 2026 · similarity 0.87)
- Generative Multimodal Models are In-Context Learners (CVPR 2024 · similarity 0.87)
- Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs (ICML 2026 · similarity 0.87)
- LLaVA-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal 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.
