E1-Math-1.5B Open-source Language Model - Usable Even with a Small Budget, Effortlessly Deduce with Flexible Inference

E1 Math 1.5B

Developed by Salesforce

E1-Math-1.5B is a language model fine-tuned based on DeepSeek-R1-Distilled-Qwen-1.5B, supporting elastic reasoning and the GRPO method, suitable for budget-constrained deduction scenarios.

Large Language Model

Transformers

#Elastic Reasoning #Budget-Constrained Deduction #Mathematical Reasoning

Downloads 295

Release Time : 5/7/2025

Model Overview

This model is trained with budget-constrained deduction strategies to achieve elastic reasoning and incorporates the GRPO method, enabling adaptive reasoning during interrupted thought processes without additional training, and generalizing to unseen budget-constrained scenarios.

Model Features

Elastic Reasoning

Supports reasoning under budget constraints, adapting to varying computational resource limits.

GRPO Method

Enables adaptive reasoning during interrupted thought processes, generalizing to unseen budget-constrained scenarios without additional training.

High Performance

Demonstrates high accuracy across multiple token lengths, particularly outperforming the base model at shorter token lengths.

Model Capabilities

Mathematical Reasoning

Elastic Reasoning

Adaptive Reasoning

Use Cases

Academic Research

Mathematical Problem Solving

Used to solve complex mathematical problems, especially in resource-constrained environments.

Demonstrates high accuracy across multiple token lengths.

Education

Mathematics-Assisted Teaching

Helps students understand and solve mathematical problems, providing elastic reasoning support.

🚀 E1-Math-1.5B Language Model

E1-Math-1.5B is a fine - tuned language model that enables adaptive reasoning and effective generalization under budget constraints.

🚀 Quick Start

E1-Math-1.5B is a language model fine - tuned from DeepSeek - R1 - Distilled - Qwen - 1.5B. It is trained for Elastic Reasoning by a budget - constrained rollout strategy, integrated into GRPO. This approach teaches the model to reason adaptively when the thinking process is cut short and generalizes effectively to unseen budget constraints without additional training.

✨ Features

Model Information

Property	Details
Library Name	transformers
Datasets	agentica - org/DeepScaleR - Preview - Dataset
Base Model	deepseek - ai/DeepSeek - R1 - Distill - Qwen - 1.5B
License	cc - by - nc - 4.0

Performance (Avg@16)

Model	Tokens	Acc (%)	Tokens	Acc (%)	Tokens	Acc (%)	Tokens	Acc (%)	Tokens	Acc (%)
DeepScaleR - 1.5B	10050	41.0	1488	5.2	1904	9.6	2809	15.8	3700	22.7
E1 - Math - 1.5B	6825	35.0	1340	13.5	1799	17.5	2650	24.8	3377	27.9

💻 Usage Examples

For detailed usage, please refer to repo.

📚 Documentation

Citation

@article{xu2025scalable,
  title={Scalable Chain of Thoughts via Elastic Reasoning},
  author={Xu, Yuhui and Dong, Hanze and Wang, Lei and Sahoo, Doyen and Li, Junnan and Xiong, Caiming},
  journal={arXiv preprint arXiv:2505.05315},
  year={2025}
}

Ethical Considerations

⚠️ Important Note

This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high - risk scenarios where errors or misuse could significantly impact people's lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.

📄 License

This project is licensed under the cc - by - nc - 4.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご