|
| 1 | +# Verus Proof Synthesis Leaderboard |
| 2 | + |
| 3 | +A web-based leaderboard for tracking progress in LLM-based formal verification for Rust using Verus. |
| 4 | + |
| 5 | +## 🌐 Live Website |
| 6 | + |
| 7 | +Visit the leaderboard at: **[https://microsoft.github.io/verus-proof-synthesis/](https://microsoft.github.io/verus-proof-synthesis/)** |
| 8 | + |
| 9 | +## 📁 Directory Structure |
| 10 | + |
| 11 | +``` |
| 12 | +leaderboard/ |
| 13 | +├── index.html # Main leaderboard page |
| 14 | +├── submit.html # Submission guidelines |
| 15 | +├── about.html # About page |
| 16 | +├── css/ |
| 17 | +│ └── main.css # Stylesheet |
| 18 | +├── js/ |
| 19 | +│ └── leaderboard.js # Interactive functionality |
| 20 | +├── data/ |
| 21 | +│ ├── schema.json # JSON schema for submissions |
| 22 | +│ ├── verus-bench-results.json # Verus-Bench leaderboard data |
| 23 | +│ └── verusage-bench-results.json # VeruSAGE-Bench leaderboard data |
| 24 | +└── README.md # This file |
| 25 | +``` |
| 26 | + |
| 27 | +## 🚀 Running Locally |
| 28 | + |
| 29 | +The leaderboard is a static website that can be served by any web server: |
| 30 | + |
| 31 | +```bash |
| 32 | +# Using Python |
| 33 | +cd leaderboard |
| 34 | +python -m http.server 8000 |
| 35 | +# Visit http://localhost:8000 |
| 36 | + |
| 37 | +# Using Node.js (if you have serve installed) |
| 38 | +npx serve . |
| 39 | + |
| 40 | +# Using PHP |
| 41 | +php -S localhost:8000 |
| 42 | +``` |
| 43 | + |
| 44 | +## 📤 Submitting Results |
| 45 | + |
| 46 | +### Quick Start |
| 47 | + |
| 48 | +1. **Run evaluation** on Verus-Bench or VeruSAGE-Bench |
| 49 | +2. **Format results** according to our [JSON schema](data/schema.json) |
| 50 | +3. **Open a PR** adding your entry to the appropriate data file |
| 51 | + |
| 52 | +### Submission Format |
| 53 | + |
| 54 | +```json |
| 55 | +{ |
| 56 | + "submission_id": "your-system-model-version", |
| 57 | + "system_name": "Your System Name", |
| 58 | + "model": "LLM Model Used", |
| 59 | + "date": "YYYY-MM-DD", |
| 60 | + "results": { |
| 61 | + "solved": 135, |
| 62 | + "total": 150, |
| 63 | + "percent_solved": 90.0, |
| 64 | + "avg_time_seconds": 28.5, |
| 65 | + "avg_cost_usd": 0.25 |
| 66 | + }, |
| 67 | + "breakdown": [ |
| 68 | + {"category": "CloverBench", "solved": 11, "total": 11} |
| 69 | + ], |
| 70 | + "paper_url": "https://arxiv.org/abs/...", |
| 71 | + "code_url": "https://github.com/...", |
| 72 | + "verified": false, |
| 73 | + "notes": "Brief description" |
| 74 | +} |
| 75 | +``` |
| 76 | + |
| 77 | +### Required Fields |
| 78 | + |
| 79 | +| Field | Description | |
| 80 | +|-------|-------------| |
| 81 | +| `submission_id` | Unique identifier (e.g., "mysystem-gpt4-v1.0") | |
| 82 | +| `system_name` | Name of your proof synthesis system | |
| 83 | +| `model` | LLM model used | |
| 84 | +| `date` | Submission date (YYYY-MM-DD) | |
| 85 | +| `results.solved` | Number of tasks solved | |
| 86 | +| `results.total` | Total tasks attempted | |
| 87 | +| `results.percent_solved` | Percentage solved | |
| 88 | + |
| 89 | +### Optional Fields |
| 90 | + |
| 91 | +| Field | Description | |
| 92 | +|-------|-------------| |
| 93 | +| `results.avg_time_seconds` | Average time per task (Verus-Bench) | |
| 94 | +| `results.avg_time_minutes` | Average time per task (VeruSAGE-Bench) | |
| 95 | +| `results.avg_cost_usd` | Average cost per task in USD | |
| 96 | +| `breakdown` | Per-source/project breakdown | |
| 97 | +| `paper_url` | Link to paper | |
| 98 | +| `code_url` | Link to code repository | |
| 99 | +| `notes` | Additional information | |
| 100 | + |
| 101 | +## ✅ Verification Status |
| 102 | + |
| 103 | +Submissions are labeled with verification status: |
| 104 | + |
| 105 | +- **Verified** — Results independently reproduced by maintainers |
| 106 | + |
| 107 | +To expedite verification, please: |
| 108 | +- Provide detailed reproduction instructions |
| 109 | +- Make evaluation scripts publicly available |
| 110 | +- Include exact Verus version used |
| 111 | + |
| 112 | +## ⚠️ Rules |
| 113 | + |
| 114 | +1. **No cheating**: Submissions using `assume false`, `#[verifier::external_body]`, or other trivial solutions will be rejected |
| 115 | +2. **Use standard Verus**: Use the recommended Verus version in benchmark README |
| 116 | +3. **Report honestly**: Results should be accurate and reproducible |
| 117 | +4. **One entry per configuration**: Submit separate entries for different model/system combinations |
| 118 | + |
| 119 | +## 🛠️ Development |
| 120 | + |
| 121 | +### Modifying the Leaderboard |
| 122 | + |
| 123 | +1. Edit HTML/CSS/JS files as needed |
| 124 | +2. Test locally with a web server |
| 125 | +3. Submit a PR with your changes |
| 126 | + |
| 127 | +### Adding New Data |
| 128 | + |
| 129 | +To add a new submission: |
| 130 | + |
| 131 | +1. Edit `data/verus-bench-results.json` or `data/verusage-bench-results.json` |
| 132 | +2. Add your entry to the `submissions` array |
| 133 | +3. Ensure your entry follows the schema in `data/schema.json` |
| 134 | +4. Submit a PR |
| 135 | + |
| 136 | +## 📊 Benchmarks |
| 137 | + |
| 138 | +### Verus-Bench (150 tasks) |
| 139 | + |
| 140 | +Algorithm-level verification tasks from classic CS problems. |
| 141 | + |
| 142 | +| Source | Tasks | Description | |
| 143 | +|--------|-------|-------------| |
| 144 | +| CloverBench | 11 | Classic CS examples | |
| 145 | +| MBPP | 78 | Formal specification problems | |
| 146 | +| Diffy | 38 | Array/loop programs | |
| 147 | +| Misc | 23 | Verus tutorial examples | |
| 148 | + |
| 149 | +### VeruSAGE-Bench (849 tasks) |
| 150 | + |
| 151 | +Repository-level verification tasks from real-world systems. |
| 152 | + |
| 153 | +| Project | Code | Tasks | Domain | |
| 154 | +|---------|------|-------|--------| |
| 155 | +| Anvil | AL | 104 | Distributed Systems | |
| 156 | +| Anvil Advanced | AC | 63 | Distributed Systems | |
| 157 | +| IronKV | IR | 118 | Key-Value Store | |
| 158 | +| Memory Allocator | MA | 89 | Systems | |
| 159 | +| Node Replication | NO | 29 | Distributed Systems | |
| 160 | +| NRKernel | NR | 204 | OS Kernel | |
| 161 | +| ATMO | OS | 157 | Microkernel | |
| 162 | +| Storage | ST | 63 | Storage Systems | |
| 163 | +| Vest | VE | 22 | Serialization | |
| 164 | + |
| 165 | +## 📚 Citation |
| 166 | + |
| 167 | +```bibtex |
| 168 | +@article{autoverus, |
| 169 | + title={AutoVerus: Automated Proof Generation for Rust Code}, |
| 170 | + author={Yang, Chenyuan and Li, Xuheng and Misu, Md Rakib Hossain and others}, |
| 171 | + journal={PACMPL}, |
| 172 | + volume={9}, |
| 173 | + number={OOPSLA2}, |
| 174 | + year={2025} |
| 175 | +} |
| 176 | +
|
| 177 | +@misc{verusage, |
| 178 | + title={VeruSAGE: A Study of Agent-Based Verification for Rust Systems}, |
| 179 | + author={Yang, Chenyuan and Neamtu, Natalie and Hawblitzel, Chris and others}, |
| 180 | + year={2025}, |
| 181 | + eprint={2512.18436}, |
| 182 | + archivePrefix={arXiv} |
| 183 | +} |
| 184 | +``` |
| 185 | + |
| 186 | +## 📧 Contact |
| 187 | + |
| 188 | +For questions or issues, please open an issue on the [main repository](https://github.com/microsoft/verus-proof-synthesis/issues). |
0 commit comments