Research map / Vision-language and multimodal / Multimodal LLMs
Language Is Not All You Need: Aligning Perception with Language Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 language-only instruction tuning (ablation) | KOSMOS-1 · Flickr30k image captioning | CIDEr: 65.2 -> 67.1 | Helped |
| Add language-only instruction tuning (ablation) | KOSMOS-1 · VQAv2 | VQA accuracy: 46.7 -> 51.0 | Helped |
| Add language-only instruction tuning (ablation) | KOSMOS-1 · VizWiz | VQA accuracy: 27.9 -> 29.2 | Helped |
| Add language-only instruction tuning (ablation) | KOSMOS-1 · COCO image captioning | CIDEr: 87.6 -> 84.7 | Hurt |
| Use multimodal chain-of-thought prompting (ablation) | KOSMOS-1 · Rendered SST-2 | Accuracy: 67.1 -> 72.9 | Helped |
| Provide verbal descriptions of categories in context (ablation) | KOSMOS-1 · CUB-derived binary image classification | Accuracy: 61.7 -> 90.0 | Helped |
| Remove extracted web-page text from the prompt (ablation) | KOSMOS-1 · WebSRC | EM: 15.8 -> 3.8 | Hurt |
| Remove extracted web-page text from the prompt (ablation) | KOSMOS-1 · WebSRC | F1: 31.3 -> 10.6 | Hurt |
| Add few-shot demonstrations | KOSMOS-1 · COCO image captioning | CIDEr: 84.7 -> 99.6 -> 101.7 -> 96.7 | Mixed |
| Add few-shot demonstrations | KOSMOS-1 · VizWiz visual question answering | VQA accuracy: 29.2 -> 31.4 -> 35.3 -> 39.0 | Helped |
| Train with multimodal data alongside text | KOSMOS-1 versus matched LLM · MemoryColor | Accuracy: 61.4 -> 76.1 | Helped |
| Train with multimodal data alongside text | KOSMOS-1 versus matched LLM · RelativeSize | Accuracy: 92.7 -> 94.2 | Helped |
Closest papers by meaning
- Grounding Multimodal Large Language Models to the World (ICLR 2024 · similarity 0.93)
- LENS: Multi-level Evaluation of Multimodal Reasoning with Large Language Models (ICLR 2026 · similarity 0.91)
- Kosmos-G: Generating Images in Context with Multimodal Large Language Models (ICLR 2024 · similarity 0.91)
- TokenSwap: Benchmarking and Reducing the Modality Gap in Multimodal LLMs (NeurIPS 2026 · similarity 0.90)
- Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs (ICML 2026 · similarity 0.90)
- LION: Empowering Multimodal Large Language Model with Dual-Level Visual Knowledge (CVPR 2024 · 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.