TaxPraBen - TaxRead: leaderboard

TaxPraBen (2026) is a Chinese tax-practice benchmark spanning tax knowledge memorization, understanding and application, evaluated zero-shot. This board is exact-match accuracy in percent on tax reading comprehension; higher is better.

Metric: Exact-match accuracy (%) on TaxRead (tax reading comprehension, exact match, 1,000 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

#ModelScoreOverall rank
1Qwen 2.5 7B (TaxPraBen checkpoint unspecified)84.8
2GPT-3.5 Turbo84.3#3076
3ERNIE-3.584.1
4Grok 383.8#614
5GPT-4o83.2#754
6chatglm3-6B80.6#8290
7DeepSeek-R1-Distill-Qwen-7B70.2#7976
8Yi-6B (TaxPraBen checkpoint unspecified)69.4
9Bayling265.8
10TaxPraBen GLM-4-9B (checkpoint unspecified)62.6
11Mistral-V0.3 (TaxPraBen checkpoint unspecified)55.8
12YAYI2-30B55.4
13Atom (TaxPraBen checkpoint unspecified)53.3
14TaxPraBen DeepSeekllm (checkpoint unspecified)42.1
15ChnLLaMA3 (TaxPraBen checkpoint unspecified)29.1

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=taxpraben-taxread · How It Works · Data refreshed daily, snapshot 2026-11-22.