DMind-3-Nano-270M

---

Table of Contents

• Introduction
• 1. Model Overview
• 2. Performance Snapshot
• 3. Highlights
• 4. Why This Matters for Web3
• 5. Tool Definitions & Schemas
• 6. Repository Layout
• 7. Quickstart
  • 7.1 Model Downloads
  • 7.2 Training & Eval
• Limitations & Disclaimer
• License
• Citation
• Contact

Introduction

🛡️ DMind-3-nano: Privacy-First On-Device Crypto Intent Recognition

Inference stays on your device. Standardized function calling for wallets, DEXs, and agents.

DMind-3-nano is a small, edge-optimized language model fine-tuned for crypto wallet and DEX intent recognition using standardized function-calling protocols. It is designed to run entirely on-device, enabling privacy-preserving, low-latency intent parsing for Web3 wallets and local agents.

In the 24/7 PVP environment of DeFi, traditional Copilots are merely assistants. You need an Agent capable of independent risk assessment. DMind-3-nano was engineered to arm the individual against institutional extraction — keeping your Alpha strategies local, private, and antifragile.

🛡️ DMind-3-nano is your Shield. ⚔️ DMind-3-mini is your Spear. 🔮 DMind-3 is your Oracle.

1. Model Overview

Property	Value
Model Name	DMind-3-nano
Organization	DMindAI
Base Architecture	google/functiongemma-270m-it
Parameter Count	270M
Context Window	2048 tokens
Precision	BF16 (train)
Supported Chains	Solana, Ethereum, BSC, Base
Best Tokens	SOL, USDC, JUP, RAY, BONK, WIF, ETH, BTC, POPCAT, BOME, TRUMP
Deployment	On-device: phones, tablets, edge CPUs

Experimental notice: Highest accuracy on the token/chain set above; other assets may need further tuning. Validate outputs before transacting.

2. Performance Snapshot

(Figure 1: DMind-3-nano significantly outperforms both the untuned base model and a similarly sized general-purpose model (Qwen3-0.6B), especially in multi-turn success.)

3. Highlights

• 🔐 Privacy-first: 100% on-device intent recognition; no data leaves the device.
• 📱 Edge-optimized: 270M params; runs on phones/tablets/edge CPUs.
• 🔄 Standardized protocols: SEARCH_TOKEN / EXECUTE_SWAP with unified schemas.
• 🌐 Multi-chain: Solana, Ethereum, BSC, Base.
• 🌍 Multilingual: English + Chinese intents.
• 🤖 Agent-native: Designed for local-first wallet/agent workflows where a growing share of trading decisions and execution happen on-device.
• 📊 Training data: The final full fine-tune used 12,000+ samples in total; 60%+ comes from real trading scenarios.
• 🧾 (To our knowledge) first public vertical-domain FunctionGemma case study: An end-to-end example of fine-tuning google/functiongemma-270m-it for a real wallet/DEX intent domain.

4. Why This Matters for Web3

Standardization as a Step-Change

Web3 is composable at the protocol layer (tokens, RPCs), but still fragmented at the intent layer. Today every wallet, DEX, and agent framework invents its own "swap/search intent" schema and function-calling format. The result is high integration cost, brittle adapters, inconsistent safety guarantees, and poor ecosystem interoperability.

This work targets a transformative goal: standardize wallet intents as a small, versionable protocol between natural language and transaction builders. Concretely, DMind-3-nano enforces a minimal set of typed tools (e.g. SEARCH_TOKEN, EXECUTE_SWAP) with strict schemas and a deterministic wrapper output format.

What standardization unlocks:

• Interoperability: One protocol works across wallets/DEXs/agents; integrations become plug-and-play.
• Safety & auditability: Tool calls are structured data — easy to validate, simulate, policy-check, and display for confirmation before signing.
• Benchmarkability: Shared datasets and comparable evaluations across models and releases.
• Ecosystem scaling: New tools can be added via versioning without breaking existing clients.

DMind-3-nano is not only a model — it is a proposal for a standard protocol layer that can make wallet intelligence as interoperable as ERC-20 made tokens.

The Next Wave: Local Agents Executing Trades

We expect a large share of future Web3 activity to be agent-driven: wallets will run local copilots that continuously parse user intent, monitor context, and propose/execute transactions. In that world, "cloud-only" intelligence becomes a bottleneck and a risk:

• Privacy: Trading intent, token preferences, and behavioral signals should not be streamed to third-party servers.
• Latency & reliability: Agents must work instantly and offline (mobile, hardware wallets, poor connectivity).
• Security boundaries: Local agents can keep a tighter loop between intent → policy checks → simulation → user confirmation → signing.

This is why a small, high-accuracy on-device function-calling model is necessary infrastructure for the agent-native wallet era.

5. Tool Definitions & Schemas

To ensure interoperability, DMind-3-nano uses strict JSON schemas for tool definitions.

SEARCH_TOKEN

Used to find token metadata or address on a specific chain.

{
  "name": "SEARCH_TOKEN",
  "description": "Search for a cryptocurrency token on-chain to retrieve its metadata or address.",
  "parameters": {
    "type": "object",
    "properties": {
      "symbol": {
        "type": "string",
        "description": "The ticker symbol of the token (e.g., 'SOL', 'USDC')."
      },
      "address": {
        "type": "string",
        "description": "The specific contract address (CA) of the token, if known."
      },
      "chain": {
        "type": "string",
        "enum": ["solana", "ethereum", "bsc", "base"],
        "description": "The target blockchain network."
      },
      "keyword": {
        "type": "string",
        "description": "General search keywords (e.g., project name) if symbol/address are unclear."
      }
    },
    "required": []
  }
}

EXECUTE_SWAP

Used to construct a swap transaction intent between two assets.

{
  "name": "EXECUTE_SWAP",
  "description": "Propose a token swap transaction.",
  "parameters": {
    "type": "object",
    "properties": {
      "inputTokenSymbol": {
        "type": "string",
        "description": "Symbol of the token being sold (e.g., 'SOL')."
      },
      "inputTokenCA": {
        "type": "string",
        "description": "Contract address of the token being sold."
      },
      "outputTokenCA": {
        "type": "string",
        "description": "Contract address of the token being bought."
      },
      "inputTokenAmount": {
        "type": "number",
        "description": "Absolute amount of input token to swap."
      },
      "inputTokenPercentage": {
        "type": "number",
        "description": "Percentage of balance to swap (0.0 to 1.0), used if exact amount is not specified."
      },
      "outputTokenAmount": {
        "type": "number",
        "description": "Minimum amount of output token expected (optional/slippage related)."
      }
    },
    "required": ["inputTokenSymbol"]
  }
}

Output Format

The model outputs the function call wrapped in special tokens (standard FunctionGemma format):

<start_function_call>call:FUNCTION_NAME{key1:val1, key2:val2}<end_function_call>

Example:

User: "Search for SOL on Solana"

Model:

<start_function_call>call:SEARCH_TOKEN{symbol:"SOL", chain:"solana"}<end_function_call>

6. Repository Layout

• model/ — Experimental model weights (use at your own risk in production)
• src/ — Training/eval utilities
  • train.py (LoRA or full fine-tune)
  • evaluate.py (benchmark evaluation)
  • prepare_dataset.py (SFT-ready formatting)
  • generate_benchmark.py (100-case benchmark)
  • config.py (tools, prompts, token maps)
• data/ — Sample data
  • training_data.json (raw; open-sourced subset for reproducibility)
  • benchmark_dataset.json (eval set; includes Chinese test prompts by design)
• results/evaluation_results.json — Sample output
• run_training.sh, requirements.txt

7. Quickstart

7.1 Model Downloads

Model	Base Model	Download
DMind-3-Nano-270M	google/functiongemma-270m-it	Hugging Face Link

7.2 Training & Eval

Install:

pip install -r requirements.txt

Train (LoRA default):

python -m src.train \
  --model_path /path/to/functiongemma-270m-it \
  --dataset_path ./data/training_data.json \
  --output_dir ./runs \
  --bf16

Switch to full fine-tune: add --no-use-lora. Use --use_4bit/--use_8bit + --gradient_checkpointing for low memory.

Evaluate:

python -m src.evaluate \
  --model_path ./runs/<run>/final_model \
  --benchmark_path ./data/benchmark_dataset.json \
  --output_path ./results/eval_$(date +%Y%m%d_%H%M%S).json

Data utilities:

# Prepare SFT data
python -m src.prepare_dataset --input ./data/training_data.json --output ./data/prepared_dataset.json
# Regenerate benchmark
python -m src.generate_benchmark --output ./data/benchmark_dataset.json

Limitations & Disclaimer

• Experimental Release: This is a bold exploratory release. We do not take responsibility for any financial losses incurred from using this model in production environments.
• Token Coverage: Highest accuracy on the supported token/chain set. Other assets may need further tuning.
• Knowledge Cutoff: Specific protocol data is limited to the training cutoff. Use with RAG for real-time data.
• Legal Disclaimer: This model is an analytical tool, not a financial advisor. The output (NFA) should never be the sole basis for investment decisions.

License

• Code: MIT
• Model card intent: Apache 2.0
• Protocol specs (SEARCH_TOKEN / EXECUTE_SWAP): Public domain for maximal adoption.
• Base Models:
  • DMind-3-Nano-270M is derived from google/functiongemma-270m-it, originally licensed under the Gemma License.
• Contributions are welcome via issues/PRs.

Citation

If you use DMind-3 in your research, please cite our paper:

DMind-3: A Sovereign Edge--Local--Cloud AI System with Controlled Deliberation and Correction-Based Tuning for Safe, Low-Latency Transaction Execution
Enhao Huang, Frank Li, Tony Ling, Lowes Yang
arXiv preprint arXiv:2602.11651, 2026
[arXiv] [PDF]

@misc{huang2026dmind3,
  title={DMind-3: A Sovereign Edge--Local--Cloud AI System with Controlled Deliberation and Correction-Based Tuning for Safe, Low-Latency Transaction Execution},
  author={Huang, Enhao and Li, Frank and Ling, Tony and Yang, Lowes},
  journal={arXiv preprint arXiv:2602.11651},
  year={2026}
}

Contact

For questions or support, please contact team@dmind.ai

• 🌐 Project Homepage: https://dmind.ai
• 💬 Community Discussion: Discord
• 🐦 Twitter: @dmind_ai