Photonic PCFBench - Mechanism Diagnosis Paired: 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: F1. Source: arxiv.org. Saturation forecast: Not forecast. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 0.7 | #650 |
| 2 | Qwen 3 VL 8B Instruct | 0.5 | #419 |
| 3 | Qwen 2.5 VL 72B Instruct | 0.5 | #383 |
| 4 | Qwen 3.5 122B A10B | 0.48 | #184 |
| 5 | Qwen 3.5 9B | 0.43 | #391 |
| 6 | Qwen 2 VL 72B Instruct | 0.38 | #551 |
| 7 | Claude Opus 4.5 (Thinking) | 0.35 | #87 (Claude Opus 4.5) |
| 8 | Qwen 3.5 27B | 0.32 | #192 |
| 9 | Gemini 3 Flash | 0.28 | #103 |
| 10 | Claude Opus 4.5 (Non-reasoning) | 0.27 | #87 (Claude Opus 4.5) |
| 11 | Qwen 3 VL 30B A3B Instruct | 0.24 | #390 |
| 12 | Qwen-VL-Max | 0.24 | #475 |
| 13 | Claude Sonnet 4.5 (Thinking) | 0.2 | #148 (Claude Sonnet 4.5) |
| 14 | O3 | 0.19 | #132 |
| 15 | GPT-5 | 0.13 | #105 |
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-mechanism-diagnosis-paired · How It Works · Data refreshed daily, snapshot 2026-11-05.