Research map / Vision-language and multimodal / Multimodal LLMs
Merlin: Empowering Multimodal LLMs with Foresight Minds
En Yu, Liang Zhao, YANA WEI, Jinrong Yang, Dongming Wu, Lingyu Kong, Haoran Wei, Tiancai Wang, Zheng Ge, Xiangyu Zhang, Wenbing Tao
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 Foresight Pre-Training data (ablation) | Merlin · MMBench prediction reasoning (Dev) | average score: 60.7 -> 64.4 | Improved |
| add Foresight Instruction-Tuning data (ablation) | Merlin · MMBench prediction reasoning (Dev) | average score: 61.2 -> 64.4 | Improved |
| include image-text pair data in pretraining (ablation) | Merlin · MMBench prediction reasoning (Dev) | average score: 52.8 -> 64.4 | Improved |
| add Trajectory Chain-of-Thought in FIT (ablation) | Merlin · MMBench prediction reasoning (Dev) | Prediction Reasoning: 63.0 -> 64.4 | Improved |
| use 448 instead of 336 input resolution (ablation) | Merlin · Prediction Rea. / GOT10k | Prediction Reasoning: 59.8 -> 64.4 | Improved |
| use Conv2d projector instead of MLP (fewer tokens) (ablation) | Merlin · Prediction Rea. / GOT10k | Prediction Reasoning: 58.1 -> 64.4 | Improved |
| unfreeze visual encoder during pre-training (ablation) | Merlin · Prediction Rea. / GOT10k | GOT10k AO: 28.4 -> 51.4 | Improved |
| add precise task description to prompts (ablation) | Merlin · GOT10k | AO: 28.4 -> 51.4 | Improved |
| add foresight learning (FPT and FIT) | LLaVA-1.5-style MLLM · POPE COCO Random | accuracy: 83.29 -> 91.58 | Improved |
| foresight learning versus LLaVA-1.5 on GQA | Merlin · GQA | accuracy: 62.0 -> 60.5 | Hurt |
Later papers that compared with this method
4 later accepted papers report Merlin as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Comprehensive language–image pre-training for 3D medical image understanding (ECCV 2026 · better on 3 of 3 shared results)
- Large-scale and Fine-grained Vision-language Pre-training for Enhanced CT Image Understanding (ICLR 2025 · better on 7 of 7 shared results)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · better on 2 of 2 shared results)
- Boosting Vision Semantic Density with Anatomy Normality Modeling for Medical Vision-language Pre-training (ICCV 2025 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FutureOmni: Evaluating Future Forecasting from Omni-Modal Context for Multimodal LLMs (ICML 2026 · similarity 0.85)
- See Further, Think Deeper: Advancing VLM's Reasoning Ability with Low-level Visual Cues and Reflection (CVPR 2026 · similarity 0.84)
- How Far Ahead Do LLMs Plan? Uncovering the Latent Horizon in Chain-of-Thought Reasoning (ICML 2026 · similarity 0.84)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · similarity 0.83)
- FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-and-Language Navigation (CVPR 2026 · similarity 0.83)
- CoFFT: Chain of Foresight-Focus Thought for Visual Language Models (NeurIPS 2025 · similarity 0.83)
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.
