AI Paper Atlas

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

How Far Are LLMs from Professional Poker Players? Revisiting Game-Theoretic Reasoning with Agentic Tool Use

Minhua Lin, Enyan Dai, Hui Liu, Xianfeng Tang, Yuliang Yan, Zhenwei Dai, Jingying Zeng, Zhiwei Zhang, Fali Wang, Hongcheng Gao, Chen Luo, Xiang Zhang and 2 more

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 2 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 solver tool calling with BC and RL (ToolPoker) over BC-RIRLQwen2.5-7B · Leduc Hold'em avgnet chip gain: -40.5 -> +6.8Improved
add solver tool calling (ToolPoker) over BC-RIRLQwen2.5-7B · Limit Texas Hold'em avgnet chip gain: -77.6 -> +45.0Improved
add behavior cloning stage before RIRL (ablation)Qwen2.5-7B · Leduc Hold'em avg (Table 2)net chip gain: -23.6 -> -9.8Improved
replace sparse outcome reward (BC-SPRL) with regret-based step reward (BC-RIRL)Qwen2.5-7B · Leduc Hold'em avg (Table 2)net chip gain: -72.6 -> -9.8Improved
RL fine-tuning variants vs vanilla on KuhnQwen2.5-7B · Kuhn Poker avgnet chip gain: -28.8 -> +17.1Improved
scale model size from 3B to 72BQwen2.5 · Leduc Hold'em avgnet chip gain: -135.5 -> -9.1Improved
add BC and RIRL fine-tuning to improve factual alignmentQwen2.5-7B · Leduc Hold'em reasoningFA: 0.86 -> 1.06No clear effect
add BC and RIRL fine-tuning to improve heuristic reasoningQwen2.5-7B · Leduc Hold'em reasoningHR: 0.95 -> 1.93Improved
add BC stage to ToolPoker (ablation)ToolPoker · Leduc Hold'em gameplay and reasoningchip gain, HRImproved
add RL fine-tuning to ToolPoker (ablation)ToolPoker · Leduc Hold'em gameplay and reasoningchip gain, FA, ACImproved
judge with GPT-5 instead of GPT-4.1-miniToolPoker · Limit Texas Hold'em reasoningLLM-judge average: 1.94 -> 1.91No clear effect
large closed LLM exploiting non-equilibrium opponents vs CFR+GPT-4o · Leduc Hold'emnet chip gain: +41.5 -> -8Mixed

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