ESF-Bench: leaderboard
ESF-Bench (ServiceNow, 2026) tests enterprise slot filling: models read a schema, a multi-turn conversation and auxiliary sources and must fill every slot of the final dialogue state. Final Joint Goal Accuracy is the share of 810 samples with all slots correct, in percent; higher is better.
Metric: Final Joint Goal Accuracy (%; share of samples whose final dialogue state has every slot correct; 810 synthetic multi-turn enterprise slot-filling samples (6,530 slots, 8 domains) whose prompts combine a schema, a user-assistant conversation and auxiliary sources; varied output formats mapped to a standard form before exact comparison with the ground truth; 32k maximum tokens, reasoning setting as printed per model). Source: arxiv.org. Saturation forecast: Around October 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 33.3 | #453 |
| 2 | GPT-5.1 (High) | 30.7 | #229 (GPT-5.1) |
| 3 | GPT-OSS-120B (High) | 23.8 | #700 (GPT-OSS-120B) |
| 4 | Qwen 3 32B (Thinking) | 22 | #960 (Qwen 3 32B) |
| 5 | Ministral-3-14B-Reasoning-2512 | 20.9 | #1396 |
| 6 | GPT-OSS-120B (Low) | 20.7 | #700 (GPT-OSS-120B) |
| 7 | GPT-5.1 (Non-reasoning) | 10.9 | #229 (GPT-5.1) |
| 8 | Gemini 2.5 Flash (Non-reasoning) | 9.2 | #453 (Gemini 2.5 Flash) |
| 9 | Qwen 3 32B (Non-reasoning) | 4.3 | #960 (Qwen 3 32B) |
| 10 | Ministral-3-14B-Instruct-2512 | 3.7 | #1593 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=esf-bench · How It Works · Data refreshed daily, snapshot 2026-11-13.