ESF-Bench - Conditionals: leaderboard

Metric: Tagged slot accuracy (%; slots tagged with a COND (Conditionals) scenario, slot values that depend on stated conditions; 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 December 2026. 10 models tracked.

Top models

#ModelScore
1GPT-5.1 (High)88.4
2Gemini 2.5 Flash88.1
3GPT-OSS-120B (High)87.2
4GPT-OSS-120B (Low)85.5
5Qwen 3 32B (Thinking)75
6Ministral-3-14B-Reasoning-251274.4
7GPT-5.1 (Non-reasoning)64.2
8Gemini 2.5 Flash (Non-reasoning)60.7
9Qwen 3 32B (Non-reasoning)59
10Ministral-3-14B-Instruct-251254.7

Interactive version: theaggregate.ai/benchmark?slug=esf-bench-conditionals · How It Works · Data refreshed daily, snapshot 2026-09-29.