AI Paper Atlas

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

From Winning to Understanding: A Diagnostic Long-Horizon RTS Benchmark for LLMs

Jiacheng Li, Jiahui Liu, Yuqing Wang, Gaochen Cui, Xiao Zhang, Qianchuan Zhao, Ziyou Zhang, Chenghao Li

ICML 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents

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 structured strategic guidance and full state description to the prompt (ablation)GPT-4o · Prompt sensitivity vs rule-based AIwin rate: 0.15 / 0.85 / 0 -> 0 / 0.59 / 0.41 (W/D/L, standard vs simplified)Improved
replace precise coordinates with ambiguous directional cuesDeepSeek · Track C Encirclementtask score: 0.600 -> 0.221Hurt
replace precise coordinates with ambiguous directional cuesClaude · Track C Encirclementtask score: 0.647 -> 0.746No clear effect
replace precise coordinates with ambiguous directional cuesGemini · Track C Encirclementtask score: 0.386 -> 0.357Hurt
replace precise coordinates with ambiguous directional cuesKimi · Track C Encirclementtask score: 0.495 -> 0.380Hurt
evaluate in cross-play against adaptive LLM opponents instead of scripted rule-based AIGPT-4o · Track A vs Track B rankingranking: win rate 0.45 (Track A, top macro throughput) -> Elo 1602.90 (third)Hurt
evaluate in cross-play against adaptive LLM opponents instead of scripted rule-based AIClaude · Track A vs Track B rankingranking: win rate 0.70 -> Elo 2136.02Improved
add the Scouting metric to a four-metric regression on win rate in Track Bwin-rate regression · Track Bstandardized beta: Scouting 0.238 vs MU -0.152Improved
use MU alone as predictor of win rate versus jointly with KLR, MC and Scoutingwin-rate regression · Track Acorrelation: r = 0.95 -> beta = 0.012Mixed
increase number of sampled games per pair in subsampling from 2 to 3Track B ranking · Track B subsamplingexact ranking recovery: 80.8% -> 93.5%Improved

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