Research map / Large language models / Agents, code and math
Evolving Quantitative Reasoning through Self-Play in Digital Twin Markets
Tianmi Ma, Wenxin Huang, Jiawei Du, Lin Li, Xian Zhong, Joey Tianyi Zhou
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 changed | Where | Measured | Effect |
|---|---|---|---|
| decouple planning from computation: add tool creation and market analysis around GPT-5-mini (DeMAgent) | GPT-5-mini · NASDAQ | Total return TR: 1.20 -> 8.49 | Improved |
| add DeMAgent pipeline to GPT-5-mini | GPT-5-mini · CSI | Total return TR: 20.05 -> 41.32 | Improved |
| add DeMAgent pipeline to GPT-5-mini | GPT-5-mini · CRYPTOCURRENCY | Total return TR: 4.21 -> 25.21 | Improved |
| add knowledge & tool creation only (ablation) | GPT-5-mini · CRYPTOCURRENCY | Total return TR: 5.19 -> 20.96 | Improved |
| add market analysis only (ablation) | GPT-5-mini · CRYPTOCURRENCY | Total return TR: 5.19 -> 9.36 | Improved |
| add market analysis only (ablation) | GPT-5-mini · NASDAQ | Win rate WR: 57.14 -> 52.38 | Hurt |
| combine market analysis with tool creation (ablation) | DeMAgent · CSI | Total return TR: 19.46 -> 29.64 | Improved |
| add evaluation & refinement on top of MA + KTC (ablation) | DeMAgent · CSI | Total return TR: 29.64 -> 41.32 | Improved |
| add evaluation & refinement on top of MA + KTC (ablation) | DeMAgent · CRYPTOCURRENCY | Total return TR: 17.30 -> 25.21 | Improved |
| add evaluation & refinement on top of MA + KTC (ablation) | DeMAgent · NASDAQ | Win rate WR: 45.24 -> 59.52 | Improved |
| use a stronger LLM backbone in DeMAgent (GPT-5-mini vs LLaMa-3.1) (ablation) | DeMAgent · NASDAQ | Total return TR: -1.59 -> 8.49 | Improved |
| wrap standalone LLMs with DeMAgent across backbones | Gemini-3-Pro · DecoupledMarket | Max drawdown MD: -1.1459 -> 0.0000 | Improved |
Closest papers by meaning
- Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning (ICLR 2026 · similarity 0.88)
- GTBench: Uncovering the Strategic Reasoning Capabilities of LLMs via Game-Theoretic Evaluations (NeurIPS 2024 · similarity 0.88)
- Non-myopic Generation of Language Models for Reasoning and Planning (ICLR 2025 · similarity 0.88)
- Win Fast or Lose Slow: Balancing Speed and Accuracy in Latency-Sensitive Decisions of LLMs (NeurIPS 2025 · similarity 0.88)
- LLM Strategic Reasoning: Agentic Study through Behavioral Game Theory (NeurIPS 2025 · similarity 0.87)
- SPIRAL: Symbolic LLM Planning via Grounded and Reflective Search (AAAI 2026 · similarity 0.87)
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.
