ESF-Bench - Long and Multi-Value: leaderboard

Metric: Tagged slot accuracy (%; slots tagged with a LMV (Long and Multi-Value) scenario, long identifiers and slots that hold several values; 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)81.5
2GPT-5.1 (High)78.6
3Gemini 2.5 Flash77.7
4GPT-OSS-120B (Low)76.3
5Ministral-3-14B-Reasoning-251275.7
6Qwen 3 32B (Thinking)74.3
7GPT-5.1 (Non-reasoning)68.5
8Gemini 2.5 Flash (Non-reasoning)63.5
9Qwen 3 32B (Non-reasoning)59.2
10Ministral-3-14B-Instruct-251252.6

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