AI Paper Atlas

Research map / Large language models / Agents, code and math

TaxReasoning: Benchmarking Knowledge-Intensive Mathematical Reasoning with Evolving Tax Laws

Nan Hu, Yike Wu, Jiaye Li, HuiKang Hu, Guilin Qi, Songlin Zhai, Yongrui Chen, Tianxing Wu, Tongtong Wu, Jiaoyan Chen, Jeff Z. Pan

AAAI 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add standard retrieval of tax documents to CoTGemini 2.5 Pro · TaxReasoningaccuracy (%): 31.37 -> 57.22Improved
add oracle gold documents to CoTGPT-4o · TaxReasoningaccuracy (%): 12.75 -> 14.76No clear effect
add standard retrieval to CoTQwen2.5-72B · TaxReasoningaccuracy (%): 17.65 -> 21.51Improved
use PoT instead of CoT promptingDeepSeek-R1 · TaxReasoningaccuracy (%): 42.16 -> 38.00Hurt
use PoT instead of CoT promptingOpenAI o3 · TaxReasoningaccuracy (%): 19.61 -> 21.57Improved
summarize retrieved documents instead of standard retrievalGemini 2.5 Pro · TaxReasoningaccuracy (%): 57.22 -> 56.33Hurt
use InlineSearch instead of Interact multi-turn retrievalDeepSeek-R1 · TaxReasoningaccuracy (%): 53.75 -> 59.48Improved
use multi-turn retrieval (Interact) instead of single-turn StandardGPT-4o · TaxReasoningaccuracy (%): 11.67 -> 13.01Mixed
use Interact instead of Standard RAGQwen2.5-72B · TaxReasoningaccuracy (%): 21.51 -> 17.14Hurt
retrieve from a structured tax knowledge graph instead of raw documentsDeepSeek-R1 and Gemini 2.5 Pro · TaxReasoningaccuracy (%)Improved
give the model open-book human access vs closed-book (humans)human CPAs · TaxReasoning (50-question sample)accuracy (%): 61.00 -> 82.00Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy