AI Paper Atlas

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

Can LLMs Identify Tax Abuse?

Andrew Blair-Stanek, Nils Holzenberger, Benjamin Van Durme

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
use the stricter 'correct' standard instead of 'viable' in analysis verificationo3 · Shelter Check analysis stepssteps passed of 193: 193/193 -> 166/193Hurt
use the stricter 'correct' standard instead of 'viable' in goal verification with analysisclaude-4 · Shelter Check goal verificationstrategies passed of 36: 36/36 -> 33/36Hurt
remove the expert analysis from the goal-verification prompt (ablation)o3 · Shelter Check goal verification (correct)strategies passed of 36: 29/36 -> 26/36Hurt
remove the expert analysis from the goal-verification prompt (ablation)claude-4 · Shelter Check goal verification (correct)strategies passed of 36: 33/36 -> 22/36Hurt
remove the expert analysis from the goal-verification prompt (ablation)gemini-2.5 · Shelter Check goal verification (correct)strategies passed of 36: 27/36 -> 31/36Improved
swap in an adversarial step designed to make the strategy failllama-3.3-70B · Shelter Check adversarial step (viable)strategies judged viable of 36: 31/36Hurt
swap in an adversarial step designed to make the strategy failclaude-4 · Shelter Check adversarial step (viable)strategies judged viable of 36: 20/36Hurt
use a reasoning model (o3, claude-4, gemini-2.5) instead of llama-3.3-70Breasoning models · Shelter Check step-clozemean grade (0-3)Improved
go from 0-shot to 1-shot promptingo3, claude-4, gemini-2.5 · Shelter Check step-clozemean grade (0-3)Improved
go from 1-shot to 2-shot promptingo3, claude-4, gemini-2.5 · Shelter Check step-clozemean grade (0-3)Hurt
use o1-preview instead of claude-3.5 for from-scratch strategy generationo1-preview · Shelter Check from-scratch generationmean expert grade (0-3): 1.23 -> 1.47Improved
use o3 rather than claude-4 as grader of from-scratch strategieso3 · Shelter Check grading vs expertSpearman rho: 0.30 -> 0.48Improved

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