Research map / Large language models / AI and society
FusionBench: A Unified Library and Comprehensive Benchmark for Deep Model Fusion
Anke Tang, Li Shen, Yong Luo, Enneng Yang, Han Hu, Lefei Zhang, Bo Du, Dacheng 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 |
|---|---|---|---|
| fine-tune the vision encoder on the target task | CLIP-ViT-B/32 · SUN397 | accuracy: 63.2 -> 75.0 | Improved |
| fine-tune the vision encoder on the target task | CLIP-ViT-B/32 · EuroSAT | accuracy: 46.0 -> 99.0 | Improved |
| fine-tune the vision encoder on the target task | CLIP-ViT-L/14 · SVHN | accuracy: 58.4 -> 97.9 | Improved |
| fine-tune on SVHN and evaluate on EuroSAT (negative transfer) | CLIP-ViT-B/32 · EuroSAT | accuracy: 46.0 -> 9.8 | Hurt |
| fine-tune on SVHN and evaluate on MNIST (positive transfer) | CLIP-ViT-B/32 · MNIST | accuracy: 48.2 -> 81.8 | Improved |
| fine-tune on SVHN and evaluate on MNIST (positive transfer) | CLIP-ViT-L/14 · MNIST | accuracy: 76.4 -> 92.2 | Improved |
| fine-tune on EuroSAT and evaluate on SVHN (negative transfer) | CLIP-ViT-L/14 · SVHN | accuracy: 58.4 -> 33.4 | Hurt |
| fine-tune a ResNet-50 on depth estimation and evaluate on segmentation | ResNet-50 · NYUv2 segmentation | mIoU: 52.0 -> 2.3 | Hurt |
| fine-tune GPT-2 on CoLA and evaluate on MRPC | GPT-2 · GLUE MRPC | accuracy: 80.4 -> 68.4 | Mixed |
Closest papers by meaning
- Deep Fusion: Efficient Network Training via Pre-trained Initializations (ICML 2024 · similarity 0.86)
- Rethinking Layer-wise Model Merging through Chain of Merges (NeurIPS 2026 · similarity 0.86)
- Towards Meta-Pruning via Optimal Transport (ICLR 2024 · similarity 0.85)
- Transformer Fusion with Optimal Transport (ICLR 2024 · similarity 0.85)
- OptMerge: Unifying Multimodal LLM Capabilities and Modalities via Model Merging (ICLR 2026 · similarity 0.84)
- CTBench: A Library and Benchmark for Certified Training (ICML 2025 · similarity 0.84)
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.
