ESF-Bench - Schema Constraints: leaderboard

Metric: Tagged slot accuracy (%; slots tagged with a SC (Schema Constraints) scenario, regex, semantic, unit and cross-slot schema constraints, defaults and normalization rules; 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-OSS-120B (High)83.8
2Gemini 2.5 Flash82.5
3GPT-OSS-120B (Low)82.4
4GPT-5.1 (High)80.7
5Ministral-3-14B-Reasoning-251277.1
6GPT-5.1 (Non-reasoning)77.1
7Qwen 3 32B (Thinking)74.6
8Gemini 2.5 Flash (Non-reasoning)70.5
9Ministral-3-14B-Instruct-251264.9
10Qwen 3 32B (Non-reasoning)62.1

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