|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
- zh |
|
|
metrics: |
|
|
- accuracy |
|
|
base_model: |
|
|
- Qwen/Qwen3-14B |
|
|
pipeline_tag: text-generation |
|
|
library_name: transformers |
|
|
tags: |
|
|
- blockchain |
|
|
- conversational |
|
|
- web3 |
|
|
- qwen3 |
|
|
eval_results: |
|
|
- task: domain-specific evaluation |
|
|
dataset: DMindAI/DMind_Benchmark |
|
|
metric: normalized web3 score |
|
|
score: 74.12 |
|
|
model: DMind-1-mini |
|
|
model_rank: 2 / 24 |
|
|
--- |
|
|
|
|
|
<p align="center"> |
|
|
<img src="figures/dmind-ai-logo.png" width="300" alt="DMind Logo" /> |
|
|
</p> |
|
|
<hr> |
|
|
<div align="center" style="line-height: 1;"> |
|
|
<a href="https://dmind.ai/" target="_blank" style="margin: 2px;"> |
|
|
<img alt="DMind Website" src="https://img.shields.io/badge/DMind-Homepage-blue?logo=data:image/svg+xml;base64,)" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
<a href="https://huggingface.co/DMindAI" target="_blank" style="margin: 2px;"> |
|
|
<img alt="Hugging Face" src="https://img.shields.io/badge/HuggingFace-DMind-ffd21f?color=ffd21f&logo=huggingface" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
<a href="https://x.com/dmind_ai" target="_blank" style="margin: 2px;"> |
|
|
<img alt="X" src="https://img.shields.io/badge/X-@DMind-1DA1F2?logo=x" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
<a href="https://huggingface.co/spaces/DMindAI/DMind-1-mini" target="_blank" style="margin: 2px;"> |
|
|
<img alt="Chat" src="https://img.shields.io/badge/🤖%20Chat-DMind--1--mini-536af5?color=536af5&logoColor=white" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
<a href="https://discord.gg/xxwmPHU3" target="_blank" style="margin: 2px;"> |
|
|
<img alt="Discord" src="https://img.shields.io/badge/Discord-DMind-7289da?logo=discord&logoColor=white&color=7289da" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
<a href="https://opensource.org/licenses/MIT" target="_blank" style="margin: 2px;"> |
|
|
<img alt="Code License: MIT" src="https://img.shields.io/badge/Code%20License-MIT-yellow.svg" style="display: inline-block; vertical-align: middle;"/> |
|
|
</a> |
|
|
</div> |
|
|
|
|
|
|
|
|
## Table of Contents |
|
|
- [Introduction](#introduction) |
|
|
- [1. Model Overview](#1-model-overview) |
|
|
- [2. Evaluation Results](#2-evaluation-results) |
|
|
- [3. Use Cases](#3-use-cases) |
|
|
- [4. Quickstart](#4-quickstart) |
|
|
- [4.1 Model Downloads](#41-model-downloads) |
|
|
- [4.2 OpenRouter API](#42-openrouter-api) |
|
|
- [4.3 OpenRouter Web Chat](#43-openrouter-web-chat) |
|
|
- [License](#license) |
|
|
- [Contact](#contact) |
|
|
|
|
|
## Introduction |
|
|
|
|
|
We introduce **DMind-1**, a domain-specialized LLM fine-tuned for the Web3 ecosystem via supervised instruction tuning and reinforcement learning from human feedback (RLHF). |
|
|
|
|
|
To support real-time and resource-constrained applications, we further introduce **DMind-1-mini**, a compact variant distilled from both DMind-1 and a generalist LLM using a multi-level distillation framework. It retains key domain reasoning abilities while operating with significantly lower computational overhead. |
|
|
|
|
|
**DMind-1** and **DMind-1-mini** represent a robust foundation for intelligent agents in the Web3 ecosystem. |
|
|
|
|
|
## 1. Model Overview |
|
|
|
|
|
### DMind-1-mini |
|
|
|
|
|
To address scenarios requiring lower latency and faster inference, we introduce **DMind-1-mini**, a lightweight distilled version of DMind-1 based on Qwen3-14B. DMind-1-mini is trained using knowledge distillation and our custom **DeepResearch** framework, drawing from two teacher models: |
|
|
- **DMind-1** (Qwen3-32B): Our specialized Web3 domain model. |
|
|
- **GPT-o3 + DeepResearch**: A general-purpose SOTA LLM, with its outputs processed through our DeepResearch framework for Web3 domain alignment. |
|
|
|
|
|
The **Distillation pipeline** combines: |
|
|
|
|
|
- **Web3-specific data distillation**: High-quality instruction-following and QA examples generated by the teacher models. |
|
|
|
|
|
- **Distribution-level supervision**: The student model learns to approximate the teachers' output distributions through soft-label guidance, preserving nuanced prediction behavior and confidence calibration. |
|
|
|
|
|
- **Intermediate representation transfer**: Knowledge is transferred by aligning intermediate representations between teacher and student models, promoting deeper structural understanding beyond surface-level mimicry. |
|
|
|
|
|
This multi-level distillation strategy enables DMind-1-mini to maintain high Web3 task performance while significantly reducing computational overhead and latency, making it suitable for real-time applications such as instant Q&A, on-chain analytics, and lightweight agent deployment. |
|
|
|
|
|
|
|
|
## 2. Evaluation Results |
|
|
|
|
|
|
|
|
 |
|
|
|
|
|
We evaluate DMind-1 and DMind-1-mini using the [DMind Benchmark](https://huggingface.co/datasets/DMindAI/DMind_Benchmark), a domain-specific evaluation suite designed to assess large language models in the Web3 context. The benchmark includes 1,917 expert-reviewed questions across nine core domain categories, and it features both multiple-choice and open-ended tasks to measure factual knowledge, contextual reasoning, and other abilities. |
|
|
|
|
|
To complement accuracy metrics, we conducted a **cost-performance analysis** by comparing benchmark scores against publicly available input token prices across 24 leading LLMs. In this evaluation: |
|
|
|
|
|
- **DMind-1** achieved the highest Web3 score while maintaining one of the lowest token input costs among top-tier models such as Grok 3 and Claude 3.7 Sonnet. |
|
|
|
|
|
- **DMind-1-mini** ranked second, retaining over 95% of DMind-1’s performance with greater efficiency in latency and compute. |
|
|
|
|
|
Both models are uniquely positioned in the most favorable region of the score vs. price curve, delivering state-of-the-art Web3 reasoning at significantly lower cost. This balance of quality and efficiency makes the DMind models highly competitive for both research and production use. |
|
|
|
|
|
|
|
|
|
|
|
## 3. Use Cases |
|
|
- **Expert-Level Question & Answering**: Provides accurate, context-aware answers on blockchain, DeFi, smart contracts, and related Web3 topics. |
|
|
- **Compliance-Aware Support**: Assists in drafting or reviewing content within regulatory and legal contexts. |
|
|
- **Content Generation in Domain**: Produces Web3-specific blog posts, documentation, and tutorials tailored to developers and users. |
|
|
- **DeFi Strategy Suggestions**: Generates insights and recommendations for yield farming, liquidity provision, and portfolio strategies based on user-provided data. |
|
|
- **Risk Management**: Suggests strategies aligned with user risk profiles for more informed decision-making in volatile markets. |
|
|
|
|
|
## 4. Quickstart |
|
|
|
|
|
### 4.1 Model Downloads |
|
|
|
|
|
| **Model** | **Base Model** | **Download** | |
|
|
|:--------------:|:--------------:|:----------------------------------------------------------------------------:| |
|
|
| DMind-1-mini | Qwen3-14B | [Hugging Face Link](https://huggingface.co/dmind-ai/dmind-1-mini) | |
|
|
|
|
|
### 4.2 OpenRouter API (Coming Soon) |
|
|
*Documentation for API access will be available soon.* |
|
|
|
|
|
### 4.3 OpenRouter Web Chat (Coming Soon) |
|
|
*Web chat interface documentation will be available soon.* |
|
|
|
|
|
## License |
|
|
- The code repository and model weights for DMind-1-mini is released under the MIT License. |
|
|
- Commercial use, modification, and derivative works (including distillation and fine-tuning) are permitted. |
|
|
- **Base Models:** |
|
|
- DMind-1-mini is derived from Qwen3-14B, originally licensed under the [Qwen License](https://github.com/QwenLM/Qwen3). |
|
|
- Please ensure compliance with the original base model licenses when using or distributing derivatives. |
|
|
|
|
|
## Contact |
|
|
For questions or support, please contact [email protected] |
