Prompts
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Text-to-Image Generation | prompts 500 sampled | HPSv20.3345 | 36 | |
| Speculative Decoding | 20 Prompts across 4 Task Categories | Mean Expected Tokens per Speculation Step6.55 | 20 | |
| Jailbreak Prompt Quality Evaluation | 500 randomly sampled prompts | Similarity81 | 16 | |
| Distribution-distance evaluation | Prompts 100 (evaluation) | Distinct-N (WM)94.1 | 14 | |
| Creative Plot Generation | 160 prompts NQD (test) | Character Development8.67 | 13 | |
| Over-generation attack | 1000 prompts (test) | Succ. @≥ 188.2 | 8 | |
| Text-to-Image Generation | prompts 10 randomly sampled | Inference Time (s)2.2322 | 6 | |
| Property-based retrieval | Prompts (test) | MAP0.48 | 6 | |
| Platform identification | 1000 prompts (held-out) | CPRSD l(f)1.1 | 5 | |
| Text-to-Video | 1,024 prompts (held-out) | VQ4.81 | 5 | |
| Panorama Generation | 14 prompts 1000 panoramas of dimensions 512x4608 | Intra-LPIPS0.58 | 4 | |
| Text-to-Image Generation | 400 prompts (test) | HPSv229.0533 | 4 | |
| Human Preference Evaluation (Harmlessness) | 1,172 Prompts (test) | Win Count (CS)677 | 3 | |
| Human Preference Evaluation (Helpfulness) | 1,172 prompts (test) | CS Wins695 | 3 | |
| Steering LLM states | 50 prompts | LogFreq (d)1.6666 | 3 | |
| 3D Scene Editing | 15 distinct single-task prompts | LLM Time10.63 | 3 | |
| LLM agent alignment evaluation | 1000 prompts (test) | Usefulness Score1 | 2 | |
| Word Count Adherence | 180 held-out prompts Length Far OOD 8–12k | Length Adherence Ratio72 | 1 | |
| Word Count Adherence | 180 held-out prompts Length Near OOD 4–8k | Length Adherence Ratio87 | 1 | |
| Word Count Adherence | 180 Prompts Length ID 1–4k (test) | Length Adherence Ratio99 | 1 | |
| Story Generation Quality | 180 held-out prompts Far OOD 8–12k | Story Quality Score44.1 | 1 | |
| Story Generation Quality | 180 prompts Near OOD 4–8k (held-out) | Story Quality Score48.2 | 1 |
Showing 22 of 22 rows