ESF-Bench - Selection: leaderboard

Metric: Tagged slot accuracy (%; slots tagged with a SEL (Selection) scenario, choosing among candidate values from the conversation, sources or schema enums; 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
1Gemini 2.5 Flash89.5
2Qwen 3 32B (Thinking)83.1
3GPT-5.1 (High)82.6
4GPT-OSS-120B (Low)81.5
5Ministral-3-14B-Reasoning-251280.8
6GPT-OSS-120B (High)80.3
7GPT-5.1 (Non-reasoning)75.6
8Gemini 2.5 Flash (Non-reasoning)72.3
9Ministral-3-14B-Instruct-251268.6
10Qwen 3 32B (Non-reasoning)64.8

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