Research map / Vision-language and multimodal / Cross-modal learning
MultiModalPFN: Extending Prior-Data Fitted Networks for Multimodal Tabular Learning
Wall Kim, Chaeyoung Song, Hanul 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 image features to fine-tuned TabPFN | TabPFN · PAD-UFES-20 | accuracy (%): 82.17 -> 85.22 | Improved |
| add image features to TabPFN on CBIS-DDSM Mass | TabPFN · CBIS-DDSM Mass | accuracy (%): 71.27 -> 74.53 | Improved |
| MMPFN vs text-specialized AutoGluon on Cloth | AutoGluon · Cloth | accuracy (%): 72.07 -> 66.26 | Hurt |
| add text to TabPFN on Cloth | TabPFN · Cloth | accuracy (%): 55.07 -> 66.26 | Improved |
| replace single-head linear projector with multi-head gated MLP (MGM) (ablation) | MMPFN · 9-dataset average | accuracy (%): 53.86 -> 57.37 | Improved |
| replace MGM with mixture-of-experts heads (ablation) | MMPFN · 9-dataset average | accuracy (%): 57.37 -> 53.23 | Hurt |
| add GLU gating to multi-head MLP (ablation) | MMPFN · 9-dataset average | accuracy (%): 55.81 -> 57.37 | Improved |
| replace FiLM with cross-attention pooler (CAP) (ablation) | MMPFN · 9-dataset average | accuracy (%): 55.48 -> 57.37 | Improved |
| add CAP at large token count (N=128) (ablation) | MGM · PAD-UFES-20 | accuracy (%): 83.91 -> 84.59 | Improved |
| add CAP when token count rises to 4x32 (ablation) | MGM · Cloth | accuracy (%): 61.76 -> 64.22 | Improved |
| increase number of non-tabular tokens with tabular inputs, no CAP (ablation) | MMPFN · PAD-UFES-20 | accuracy (%) | Hurt |
| train on only 10% of data vs TIP | TIP · PAD-UFES-20 | accuracy (%): 70.44 -> 72.87 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- TabSTAR — TabSTAR: A Tabular Foundation Model for Tabular Data with Text Fields (NeurIPS 2025) · better on 2 of 4 shared results
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
- MulTaBench: Benchmarking Multimodal Tabular Learning with Text and Image (NeurIPS 2026 · similarity 0.88)
- A Closer Look at TabPFN v2: Understanding Its Strengths and Extending Its Capabilities (NeurIPS 2025 · similarity 0.88)
- CyIN: Cyclic Informative Latent Space for Bridging Complete and Incomplete Multimodal Learning (NeurIPS 2025 · similarity 0.87)
- Amplifying Prominent Representations in Multimodal Learning via Variational Dirichlet Process (NeurIPS 2025 · similarity 0.86)
- MSP: Modality Self-Play for Training Multimodal LLMs Without Paired Cross-Modal Data (NeurIPS 2026 · similarity 0.86)
- TuneTables: Context Optimization for Scalable Prior-Data Fitted Networks (NeurIPS 2024 · 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.
