Photonic PCFBench - Geometry Description ROUGE: leaderboard
PCFBench v1 Appendix full results; 26 isolated tasks across six categories; temperature0.1, task-specific64–2048token budgets; native fractions/errors. Paired structure+field and Field-only contexts separate. Closed-loop success uses expert sign/magnitude constraints, not physical resimulation; Human Score mean of six dimensions/two experts on120samples. Source-disputed summary/appendix cells and unbound checkpoint variants withheld.
Metric: ROUGE-L. Source: arxiv.org. Saturation forecast: Around 2035. 31 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 (Thinking) | 0.19 | #87 (Claude Opus 4.5) |
| 2 | GPT-5 | 0.19 | #105 |
| 3 | Gemini 3 Flash | 0.19 | #103 |
| 4 | Claude Opus 4.5 (Non-reasoning) | 0.18 | #87 (Claude Opus 4.5) |
| 5 | Claude Sonnet 4.5 (Thinking) | 0.18 | #148 (Claude Sonnet 4.5) |
| 6 | O3 | 0.18 | #132 |
| 7 | GPT-5 Mini | 0.18 | #193 |
| 8 | Claude Sonnet 4.5 (Non-reasoning) | 0.17 | #148 (Claude Sonnet 4.5) |
| 9 | O4 Mini | 0.17 | #185 |
| 10 | GPT-4o | 0.16 | #357 |
| 11 | GPT-4.1 Mini | 0.16 | #363 |
| 12 | Gemini 2.5 Flash Lite | 0.16 | #432 |
| 13 | Qwen 3.5 27B | 0.16 | #192 |
| 14 | Qwen 2.5 VL 72B Instruct | 0.16 | #383 |
| 15 | Qwen 3.5 122B A10B | 0.16 | #184 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=photonic-pcfbench-geometry-description-rouge · How It Works · Data refreshed daily, snapshot 2026-11-05.