KoNeoBench - Source Words - F1: leaderboard
KoNeoBench evaluates Korean neologism understanding: contextual identification, component source-word recovery, semantic and specialist-domain categorization, and definition generation. Table 6 identifies the instruction-tuned checkpoints behind shortened result labels. This series reports KoNeoBench - Source Words - F1 from 2609.19916v1, S4.T1: F1 (%). Higher is better.
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Around May 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 49.96 | #125 |
| 2 | GPT-4.1 | 43.78 | #507 |
| 3 | Solar Pro 3 | 32.9 | #1802 |
| 4 | EXAONE 3.5 7.8B Instruct | 24 | #3078 |
| 5 | Qwen 3.5 9B | 23.62 | #862 |
| 6 | Qwen 2.5 7B Instruct | 18.19 | #2678 |
| 7 | Llama 3.1 8B Instruct | 15.5 | #4318 |
No result here: #4 Claude Opus 5.5, #5 GPT-6 Astra, #9 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=koneobench-source-words-f1 · How It Works · Data refreshed daily, snapshot 2026-11-08.