Ebisu - JF-TE - F1: leaderboard
Evaluates Japanese financial implicit commitment recognition (JF-ICR) and terminology extraction (JF-TE). This is the paper v1 result table, whose stated generation limit is 1,024; current repository task defaults differ. JF-ICR uses accuracy, JF-TE uses maximal-term F1 and hit rates at 1/5/10, and the mean combines five metrics. Published fractions are converted to percentages. Explicit zero scores are retained; malformed predictions can receive zero under the public evaluator. Source-specified dated checkpoints are retained. This series reports Ebisu - JF-TE - F1 using Maximal-Term F1 (%).
Metric: Maximal-Term F1 (%). Source: arxiv.org. Saturation forecast: Around February 2027. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2 | 43.96 | #252 |
| 2 | Llama 3.3 70B Instruct | 43.7 | #541 |
| 3 | Claude Sonnet 4.5 | 43.62 | #148 |
| 4 | Qwen 3 235B A22B 2507 Instruct FP8 | 43.55 | #234 |
| 5 | Qwen 2.5 32B Instruct | 41.37 | #509 |
| 6 | Llama 4 Scout | 41.33 | #648 |
| 7 | Llama-3.3-Swallow-70B-Instruct-v0.4 | 40.7 | #540 |
| 8 | GPT-4o | 39.97 | #357 |
| 9 | Ministral-3-14B-Instruct-2512 | 37.56 | #625 |
| 10 | DeepSeek V3.1 | 36.53 | #275 |
| 11 | Qwen 3 32B | 28.6 | #449 |
| 12 | Qwen 3 14B | 20.38 | #542 |
| 13 | Qwen-14B | 16.13 | #1153 |
| 14 | Qwen 3 8B | 5.48 | #678 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ebisu-jf-te-f1 · How It Works · Data refreshed daily, snapshot 2026-11-05.