Research map / Image generation / GANs
GALIP: Generative Adversarial CLIPs for Text-to-Image Synthesis
Ming Tao, Bing-Kun Bao, Hao Tang, Changsheng Xu
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 |
|---|---|---|---|
| Integrate CLIP-based discriminator and CLIP-empowered generator rather than using CLIP contrastive loss alone | Text-to-image GAN · COCO test zero-shot | ZS-FID: 26.94 -> 16.12 | Improved |
| Add CLIP-based discriminator with CLIP-FE (ablation) | One-stage text-to-image GAN · COCO test | FID: 17.31 -> 7.92 | Improved |
| Add CLIP-based discriminator with CLIP-FE (ablation) | One-stage text-to-image GAN · COCO test | CLIPSIM: 0.2996 -> 0.3221 | Improved |
| Replace CLIP-FE with CCM&CSM (ablation) | CLIP-based discriminator · COCO test | FID: 7.92 -> 10.77 | Hurt |
| Add Bridge-FP to the CLIP-based generator (ablation) | GALIP without Prompt Predictor · COCO test | FID: 7.92 -> 6.52 | Improved |
| Add Bridge-FP to the CLIP-based generator (ablation) | GALIP without Prompt Predictor · COCO test | CLIPSIM: 0.3221 -> 0.3301 | Improved |
| Add Prompt Predictor (ablation) | GALIP with Bridge-FP · COCO test | FID: 6.52 -> 5.85 | Improved |
| Add Prompt Predictor (ablation) | GALIP with Bridge-FP · COCO test | CLIPSIM: 0.3301 -> 0.3338 | Improved |
| Use CCM&CSM instead of CLIP-FE (ablation) | CLIP-based discriminator · COCO test | FID: 7.92 -> 10.77 | Hurt |
| Use CLIP-FE features from layers 2, 5, and 9 (ablation) | GALIP discriminator · COCO test | FID: 6.70 -> 5.85 | Improved |
| Add CLIP-FE feature from layer 8 (ablation) | GALIP discriminator using layers 2, 5, and 9 · COCO test | FID: 5.85 -> 6.01 | Hurt |
| Use layer 12 instead of layer 9 as the third CLIP-FE feature (ablation) | GALIP discriminator · COCO test | FID: 5.85 -> 6.61 | Hurt |
Later papers that compared with this method
1 later accepted paper reports GALIP as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Adversarial Text to Continuous Image Generation (CVPR 2024 · better on 0 of 3 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
- un$^2$CLIP: Improving CLIP's Visual Detail Capturing Ability via Inverting unCLIP (NeurIPS 2025 · similarity 0.87)
- Diffusion Feedback Helps CLIP See Better (ICLR 2025 · similarity 0.86)
- Robustness in Both Domains: CLIP Needs a Robust Text Encoder (NeurIPS 2025 · similarity 0.85)
- Demystifying CLIP Data (ICLR 2024 · similarity 0.85)
- Exploring Sparse MoE in GANs for Text-conditioned Image Synthesis (CVPR 2025 · similarity 0.85)
- IT3D: Improved Text-to-3D Generation with Explicit View Synthesis (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.
