AI Paper Atlas

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

ARBench: Algorithmic Reasoner or API Alchemist? Evaluating LLMs Beyond API Calls

Ren-Biao Liu, Chao-Zeng Ma, Anqi Li, Hui Sun, Xin-Ye Li, Ming Li

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
go from Easy to Hard tasks on ARB-Modelo4-mini · ARB-ModelPass@1: 94.29 -> 26.15Hurt
go from Easy to Hard tasks on ARB-ModelGemini-2.5-Pro-preview · ARB-ModelPass@1: 94.29 -> 29.23Hurt
switch from ARB-Human to ARB-Model splitDeepseek-V3 · ARBench AllPass@1: 64.84 -> 47.67Hurt
switch from ARB-Human to ARB-Model splitQwen2.5-Coder-1.5B-Instruct · ARBench AllPass@1: 31.25 -> 4.65Hurt
use reasoning model DeepSeek-R1 instead of DeepSeek-V3DeepSeek · ARB-ModelPass@1: 47.67 -> 59.30Improved
use reasoning model DeepSeek-R1 instead of DeepSeek-V3DeepSeek · ARBench overallPass@1: 55.00 -> 60.33Improved
scale Qwen3 from 14B to 32BQwen3 · ARBench overallPass@1: 43.67 -> 54.33Improved
scale Qwen3 from 32B to 235B-A22BQwen3 · ARBench overallPass@1: 54.33 -> 55.33Improved
scale Qwen2.5-Coder from 1.5B to 14BQwen2.5-Coder · ARBench overallPass@1: 16.00 -> 40.67Improved
use Qwen3-14B instead of Qwen2.5-Coder-14B14B model · ARBench overallPass@1: 40.67 -> 43.67Improved
use o4-mini instead of o3-miniOpenAI reasoning · ARB-Model allPass@1: 59.30 -> 65.12Improved
use o3-mini instead of o4-miniOpenAI reasoning · ARB-Human allPass@1: 67.97 -> 68.75Mixed

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