Research map / Image generation / Text-to-image diffusion
CAP: Evaluation of Persuasive and Creative Image Generation
Aysan Aghazadeh, Adriana Kovashka
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 LLM with CPO for AIM (ablation) | AIM metric · human agreement, all ads | Krippendorff alpha: 0.18 -> 0.68 | Improved |
| add component-based scoring instead of direct LLM scoring (ablation) | persuasiveness metric · human agreement, all ads | Krippendorff alpha: 0.27 -> 0.65 | Improved |
| add AIM alignment term to Pcomp (ablation) | persuasiveness metric · human agreement, all ads | Krippendorff alpha: 0.65 -> 0.80 | Improved |
| set reason weight alpha to 4 or 5 instead of 1 (ablation) | AIM metric · human agreement subset | Krippendorff alpha: 0.51 -> 0.68 | Improved |
| use C_obj instead of LLM creativity scoring | creativity metric · human agreement, all ads | Krippendorff alpha: 0.04 -> 0.54 | Improved |
| expand message with LLAMA3 before T2I (I_LLAMA3 vs I_AR) | AuraFlow · PSA ads | AIM: 0.31 -> 0.43 | Improved |
| expand message with LLAMA3 before T2I | AuraFlow · PSA ads | C_obj: 1.36 -> 1.87 | Improved |
| expand message with LLAMA3 before T2I | AuraFlow · PSA ads | P_comp+AIM: 0.42 -> 0.60 | Improved |
| expand message with LLM before T2I | T2I models · commercial and PSA ads | AIM, creativity, persuasiveness: 6%, 7%, 10% commercial; 39%, 38%, 43% PSA | Improved |
| use QwenLM instead of LLAMA3 for message expansion | AuraFlow · PSA ads | C_obj: 1.87 -> 2.05 | Improved |
| add expansion with LLAMA3 | AuraFlow · commercial ads | AIM: 0.50 -> 0.53 | Improved |
| swap MLLM or evaluation LLM (InternVL vs QwenVL, LLAMA3 vs QwenLM) (ablation) | AIM metric · human agreement, all ads | Krippendorff alpha: 0.68 -> 0.65 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ImageReward — ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation (NeurIPS 2023) · better on 2 of 2 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
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.87)
- A3: Towards Advertising Aesthetic Assessment (CVPR 2026 · similarity 0.87)
- Revisiting text-to-image evaluation with Gecko: on metrics, prompts, and human rating (ICLR 2025 · similarity 0.87)
- Evaluation of Text-to-Video Generation Models: A Dynamics Perspective (NeurIPS 2024 · similarity 0.86)
- Rich Human Feedback for Text-to-Image Generation (CVPR 2024 · similarity 0.85)
- MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models (NeurIPS 2025 · 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.
