Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Train 'n Trade: Foundations of Parameter Markets
Tzu-Heng Huang, Harit Vishwakarma, Frederic Sala
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 |
|---|---|---|---|
| add parameter trading with alignment vs separate training | ResNet20 · TinyImageNet | testing accuracy: +10.02% and +15.93% | Improved |
| add model alignment before merging (ablation) | parameter trading · MNIST + MLP | testing accuracy: 84.64% -> 86.96% | Improved |
| add model alignment before merging (ablation) | parameter trading · CIFAR10 + ResNet20 | testing accuracy: 78.31% -> 79.90% | Improved |
| add model alignment before merging (ablation) | parameter trading · TinyImageNet + ResNet20 | testing accuracy: 31.28% -> 31.69% | Improved |
| replace broker-assisted merge with FedAvg | parameter trading · CIFAR10 + ResNet20 | testing accuracy: 79.90% -> 70.35% | Hurt |
| replace broker-assisted merge with FedAvg | parameter trading · TinyImageNet + ResNet20 | testing accuracy: 31.69% -> 19.95% | Hurt |
| FedAvg vs out-of-market training | FedAvg · MNIST + MLP | testing accuracy: 68.50% -> 81.98% | Improved |
| trade only deep layers {3, 4} (ablation) | MLP parameter trading · MNIST + MLP | testing accuracy: 71.29% -> 66.28% | Hurt |
| trade shallow layers {0, 1} instead of deep layers {3, 4} (ablation) | MLP parameter trading · MNIST + MLP | testing accuracy: 66.28% -> 78.86% | Improved |
| trade all layers instead of a subset {0, 1} (ablation) | MLP parameter trading · MNIST + MLP | testing accuracy: 78.86% -> 86.96% | Improved |
| add parameter trading vs out-of-market | linear regression · synthetic | testing loss: 42.84% and 23.88% lower loss for agents A and B | Improved |
| trade parameters across more distant tasks | linear regression · synthetic different tasks | relative testing loss improvement | Mixed |
Closest papers by meaning
- A Theory of Training Profit-Optimal LLMs (NeurIPS 2026 · similarity 0.84)
- Pricing and Competition for Generative AI (NeurIPS 2024 · similarity 0.84)
- MarS: a Financial Market Simulation Engine Powered by Generative Foundation Model (ICLR 2025 · similarity 0.84)
- A Parametric Contextual Online Learning Theory of Brokerage (ICML 2025 · similarity 0.83)
- Fairshare Data Pricing via Data Valuation for Large Language Models (NeurIPS 2025 · similarity 0.83)
- A Market for Accuracy: Classification Under Competition (ICML 2025 · similarity 0.82)
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.
