TaxPraBen - TaxCrime: leaderboard
TaxPraBen (2026) is a Chinese tax-practice benchmark spanning tax knowledge memorization, understanding and application, evaluated zero-shot. This board is accuracy in percent on identifying the tax law that applies to a case; higher is better.
Metric: Accuracy (%) on TaxCrime (tax law identification, 200 items) in TaxPraBen, zero-shot, Chinese prompts written and rated by tax-domain annotators; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | ERNIE-3.5 | 38.5 | |
| 2 | TaxPraBen GLM-4-9B (checkpoint unspecified) | 31.5 | |
| 3 | Grok 3 | 29 | #614 |
| 4 | GPT-4o | 13.5 | #754 |
| 5 | YAYI2-30B | 12.5 | |
| 6 | Qwen 2.5 7B (TaxPraBen checkpoint unspecified) | 10 | |
| 7 | DeepSeek-R1-Distill-Qwen-7B | 8 | #7976 |
| 8 | Yi-6B (TaxPraBen checkpoint unspecified) | 8 | |
| 9 | Gemma (TaxPraBen checkpoint unspecified) | 3.5 | |
| 10 | internlm2.5-7B (TaxPraBen checkpoint unspecified) | 3 | |
| 11 | Atom (TaxPraBen checkpoint unspecified) | 3 | |
| 12 | TaxPraBen DeepSeekllm (checkpoint unspecified) | 3 | |
| 13 | GPT-3.5 Turbo | 2.5 | #3076 |
| 14 | Baichuan2 (TaxPraBen checkpoint unspecified) | 2.5 | |
| 15 | ChnLLaMA3 (TaxPraBen checkpoint unspecified) | 1.5 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=taxpraben-taxcrime · How It Works · Data refreshed daily, snapshot 2026-11-22.