Research map / Vision-language and multimodal / Cross-modal learning
BIP: Bi-level Information Transfer and Completion Prompting for Visual Recognition with Missing Modalities
Haoran Fan, Xu Han, Xianglong Bao, Zheng Gao, Qi Fan, Yang Song, Jiaojiao Jiang
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 missing-type conditioning to HyperPrompt (ablation) | HyperPrompt · MM-IMDb | F1-Macro: 53.79 -> 54.32 | Improved |
| replace DCP correlated prompt with missing-type HyperPrompt | DCP · MM-IMDb | F1-Macro: 53.41 -> 54.32 | Improved |
| add missing-type conditioning (CKA, Layer6) (ablation) | HyperPrompt · MM-IMDb TXT-missing | CKA similarity (%): 31.13 -> 52.01 | Improved |
| add missing-type conditioning (CKA, Layer4) (ablation) | HyperPrompt · MM-IMDb TXT-missing | CKA similarity (%): 68.34 -> 77.99 | Improved |
| add low-rank instance alignment (ablation) | AlignPrompt · MM-IMDb | F1-Macro: 53.89 -> 54.32 | Improved |
| add MI completion (ablation) | AlignPrompt · MM-IMDb | F1-Macro: 53.47 -> 54.32 | Improved |
| add both alignment and MI completion (ablation) | AlignPrompt · MM-IMDb | F1-Macro: 53.01 -> 54.32 | Improved |
| add HyperPrompt to AlignPrompt (ablation) | BIP · MM-IMDb (50% image missing rate) | F1-Macro: 55.89 -> 57.39 | Improved |
| prompt completion instead of feature reconstruction (Knowledge Bridger) | AlignPrompt · MM-IMDb (50% image missing rate) | F1-Macro: 51.95 -> 55.89 | Improved |
| prompt completion instead of feature reconstruction | AlignPrompt · MM-IMDb image-missing validation | logit KL to full-modality prediction: 75.5% of samples lower KL | Improved |
| replace DCP with BIP on tri-modal data (double-missing) | BIP · CMU-MOSI double-missing | ACC: 70.14 -> 72.96 | Improved |
| increase trainable parameters from 2.8M to 7.9M (ablation) | BIP · MM-IMDb test | F1-Macro (%): 53.18 -> 54.32 | Improved |
Closest papers by meaning
- Deep Correlated Prompting for Visual Recognition with Missing Modalities (NeurIPS 2024 · similarity 0.93)
- Retrieval-Augmented Dynamic Prompt Tuning for Incomplete Multimodal Learning (AAAI 2025 · similarity 0.86)
- Exploring the Transferability of Visual Prompting for Multimodal Large Language Models (CVPR 2024 · similarity 0.86)
- MmAP: Multi-Modal Alignment Prompt for Cross-Domain Multi-Task Learning (AAAI 2024 · similarity 0.86)
- Delving into Multimodal Prompting for Fine-Grained Visual Classification (AAAI 2024 · similarity 0.86)
- Token-Level Contrastive Learning with Modality-Aware Prompting for Multimodal Intent Recognition (AAAI 2024 · similarity 0.85)
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.
