---

---

Nexis Labs

Pioneering the Future of Agentic AI & Blockchain Intelligence

Nexis Labs is the research and engineering organization behind Nex-T1, an enterprise-grade AI platform engineered to bridge the gap between advanced large language models (LLMs) and actionable blockchain execution. We specialize in building autonomous agents, intelligent orchestration layers, and secure human-in-the-loop (HITL) systems that operate with high reliability in decentralized environments.

---

Nex-T1: Multi-Agent Orchestration for Autonomous DeFi Trading

Official Implementation of the research paper: "Nex-T1: A Multi-Agent Orchestration Framework for Autonomous Decentralized Finance Trading" Nexis Labs Research Team, October 2025

📋 Table of Contents

• Overview
• Research Paper
• Key Results
• System Architecture
• Installation
• Reproducibility
• Experiments
• Results & Analysis
• Documentation
• Citation
• License & Usage
• Disclaimer
• Contact

---

🎯 Overview

Nex-T1 is a novel multi-agent orchestration framework designed for autonomous decentralized finance (DeFi) trading. The system addresses critical challenges in DeFi markets including:

• ⚡ Oracle Latency - Delayed price feeds creating execution risk
• 🎯 Maximal Extractable Value (MEV) - Front-running and sandwich attacks
• 💧 Slippage Management - Price impact in automated market makers
• 🌉 Cross-Chain Complexity - Routing across EVM and Solana ecosystems

Core Innovation

Nex-T1 employs a hierarchical multi-agent architecture with 25 specialized agents organized into coordinated teams:

• Research Team (7 agents): Bull/Bear/Neutral researchers, fundamental/sentiment/technical/news analysts
• Risk Management Team (5 agents): Risk Governor, Oracle Validator, MEV Monitor, Position Limiter, Compliance Auditor
• Execution Team (10 agents): EVM/Solana routers, gas optimizers, custody managers, transaction builders
• Governance & Memory Team (2 agents): Memory Curator (RAG), Governance Arbiter (voting)

Each agent is powered by GPT-4 Turbo and augmented with Retrieval-Augmented Generation (RAG) using Pinecone vector database (1536-dimensional embeddings, cosine similarity) for context-aware decision making.

---

📄 Research Paper

Paper Details

• Title: Nex-T1: A Multi-Agent Orchestration Framework for Autonomous Decentralized Finance Trading
• Authors: Nexis Labs Research Team
• Published: arXiv preprint (October 2025)
• arXiv ID: arXiv:2510.XXXXX
• Categories: cs.AI (Artificial Intelligence), cs.LG (Machine Learning), q-fin.TR (Trading and Market Microstructure)
• Pages: 23 (20 main text + 3 appendices)
• PDF: paper/main.pdf
• LaTeX Source: paper/

Abstract

Decentralized Finance (DeFi) trading presents unique challenges including oracle latency, maximal extractable value (MEV), slippage, and cross-chain execution complexity. We introduce Nex-T1, a novel multi-agent orchestration framework that leverages large language models (LLMs) and specialized agent roles to autonomously execute DeFi trading strategies across Ethereum Virtual Machine (EVM) and Solana ecosystems. Our system employs a hierarchical architecture with 25 specialized agents organized into research, risk management, execution, and governance teams. Each agent is equipped with retrieval-augmented generation (RAG) capabilities using vector memory (Pinecone, 1536-dimensional embeddings) for context-aware decision making. We formalize the multi-objective optimization problem balancing returns against slippage, MEV exposure, and oracle manipulation risks. Experimental results on historical and simulated trading scenarios demonstrate that Nex-T1 achieves a Sharpe ratio of 2.34 (vs. 1.42 for TradingAgents baseline), reduces execution costs by 31%, and maintains drawdown below 12% while executing across Uniswap V3 (Base chain) and Jupiter (Solana).

Key Contributions

1. Hierarchical Multi-Agent Architecture: Formal governance protocols with state machines, task handoffs, and majority-voting consensus for decision conflicts
2. DeFi-Specific Risk Management: Mathematical models for slippage, oracle delay, and MEV exposure with safety bounds and validation gates
3. Cross-Chain Execution Orchestration: Routing across EVM (Base L2, Uniswap V3) and Solana (Jupiter aggregator) with sub-3-second median latency
4. Vector Memory Integration: RAG with Pinecone (3 namespaces: market-intel, trade-history, agent-knowledge) improving decision quality by 27%
5. Comprehensive Empirical Validation: 6-month historical backtest with baselines, ablations, and rigorous statistical analysis

---

🏆 Key Results

Performance Comparison (6-Month Backtest: Jan-Jun 2025)

Improvements Over Baseline

• Sharpe Ratio: +65% improvement (2.34 vs. 1.42)
• Execution Cost: -31% reduction ($0.08 vs. $0.116)
• Maximum Drawdown: -47% reduction (12.3% vs. 23.1%)
• Slippage: -40% reduction (0.41% vs. 0.68%)
• Sortino Ratio: +65% improvement (3.12 vs. 1.89)

Ablation Studies

Impact of removing key components:

Key Insight: All components (multi-agent collaboration, RAG, risk management, cross-chain routing) contribute significantly to performance.

---

🏗️ System Architecture

Hierarchical Agent Organization

graph TD
    User[User Request] --> Supervisor[Supervisor Agent]
    Supervisor --> Research[Research Team]
    Supervisor --> Risk[Risk Management Team]
    Supervisor --> Execution[Execution Team]
    Supervisor --> Governance[Governance & Memory Team]

    subgraph Research [Research Team]
        Bull[Bull Researcher]
        Bear[Bear Researcher]
        Neutral[Neutral Researcher]
        Fund[Fundamental Analyst]
        Sent[Sentiment Analyst]
        Tech[Technical Analyst]
        News[News Analyst]
    end

    subgraph Risk [Risk Management Team]
        Gov[Risk Governor]
        Oracle[Oracle Validator]
        MEV[MEV Monitor]
        Pos[Position Limiter]
        Audit[Compliance Auditor]
    end

    subgraph Execution [Execution Team]
        EVM[EVM Router]
        SOL[Solana Router]
        Gas[Gas Optimizer]
        Custody[Custody Manager]
        Build[Transaction Builder]
    end

    subgraph Governance [Governance & Memory]
        Mem[Memory Curator]
        Arbiter[Governance Arbiter]
    end

Agent Communication Protocol

Agents communicate via structured JSON contracts:

SupervisorTask:

{
  "task_id": "uuid",
  "type": "research" | "execute" | "risk_check",
  "payload": {...},
  "deadline": "ISO-8601 timestamp",
  "priority": 0-10
}

TradeInput (EVM):

{
  "chain_id": 8453,
  "token_in": "0x...",
  "token_out": "0x...",
  "amount_in": "1.5 ETH",
  "slippage_bps": 50,
  "deadline_sec": 300,
  "wallet_address": "0x..."
}

RiskReport:

{
  "approved": true,
  "risk_score": 0.0-1.0,
  "checks": {
    "slippage": "PASS",
    "oracle_deviation": "PASS",
    "mev_exposure": "WARN",
    "position_limit": "PASS"
  },
  "veto_reason": null
}

Governance & Consensus

Majority Voting Mechanism: For critical decisions (trade execution), Supervisor collects votes from Research agents (Bull, Bear, Neutral). Consensus requires >50% agreement. Ties broken by Risk Governor veto.

Theorem (Multi-Agent Voting Safety): If at least ⌈n/2⌉ + 1 agents vote to reject a trade, the trade is rejected, ensuring no single malicious agent can force execution of risky trades.

Vector Memory (RAG)

Pinecone Namespaces:

Retrieval: For query q, embed to e_q ∈ ℝ^1536, retrieve top-k=5 by cosine similarity, augment agent prompt with retrieved context.

---

🚀 Installation

Prerequisites

• Python: 3.11 or higher
• Operating System: Linux, macOS, or Windows (WSL recommended)
• Memory: 8 GB RAM minimum (16 GB recommended)
• Storage: 2 GB free space

Step 1: Clone Repository

git clone https://github.com/Nexis-AI/Nex-T1-Research.git
cd Nex-T1-Research

Step 2: Create Virtual Environment

# Using venv
python3.11 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Or using conda
conda create -n nex-t1 python=3.11
conda activate nex-t1

Step 3: Install Dependencies

pip install -r requirements.txt

Key Dependencies:

• openai>=1.0.0 - GPT-4 Turbo API access
• pinecone-client>=3.0.0 - Vector database for RAG
• web3>=6.0.0 - Ethereum/EVM interactions
• solana>=0.30.0 - Solana blockchain interactions
• pandas>=2.0.0, numpy>=1.24.0 - Data processing
• matplotlib>=3.7.0, seaborn>=0.12.0 - Visualization
• pytest>=7.4.0 - Testing framework

Step 4: Configure Environment

cp .env.example .env

Edit .env with your API keys:

# Required
OPENAI_API_KEY=sk-...                    # OpenAI API key
PINECONE_API_KEY=...                     # Pinecone API key
PINECONE_ENVIRONMENT=us-west1-gcp        # Pinecone region

# Optional (for live trading)
BASE_RPC_URL=https://base-mainnet.g.alchemy.com/v2/...
SOLANA_RPC_URL=https://api.mainnet-beta.solana.com
THIRDWEB_SECRET_KEY=...                  # EVM custody
COINBASE_CDP_API_KEY=...                 # Solana custody

# Optional (for data)
CRYPTOCOMPARE_API_KEY=...                # Price data
DUNE_API_KEY=...                         # On-chain analytics

Step 5: Verify Installation

python -c "import openai, pinecone; print('Installation successful!')"

---

🔬 Reproducibility

All experiments from the paper are fully reproducible with provided code, data, and configuration.

Reproducibility Checklist

• ✅ Code: Complete implementation in src/ and experiments/
• ✅ Data: Historical OHLCV data (Jan-Jun 2025) via CryptoCompare API
• ✅ Seeds: Fixed random seeds (42, 123, 456) for all experiments
• ✅ Hardware: AWS EC2 m5.2xlarge (8 vCPU, 32 GB RAM) specifications provided
• ✅ LLM: GPT-4 Turbo (gpt-4-turbo-2024-04-09) with temperature=0.7
• ✅ Hyperparameters: All settings documented in experiments/configs/
• ✅ Dependencies: Pinned versions in requirements.txt

Data Sources

Historical Price Data:

• Source: CryptoCompare API (public, free tier available)
• Assets: ETH/USDC, SOL/USDC
• Timeframe: January 1 - June 30, 2025
• Frequency: 1-hour OHLCV (Open, High, Low, Close, Volume)
• Script: data/fetch_historical_data.py

On-Chain Data:

• Source: Dune Analytics (public queries)
• Metrics: DEX volumes, liquidity depths, gas prices
• Script: data/fetch_onchain_data.py

Sentiment Data:

• Source: CryptoCompare News API
• Coverage: Crypto news articles, social sentiment scores
• Script: data/fetch_sentiment_data.py

Experimental Setup

Backtest Configuration:

# experiments/configs/config_base.yaml
trading_pairs:
  - ETH/USDC
  - SOL/USDC

chains:
  - base  # EVM (Uniswap V3)
  - solana  # Jupiter aggregator

timeframe:
  start: 2025-01-01
  end: 2025-06-30
  frequency: 4h  # Rebalance every 4 hours

risk_parameters:
  slippage_cap_bps: 50  # 0.5%
  oracle_deviation_max: 0.01  # 1%
  position_limit_pct: 0.01  # Max 1% of pool depth

rag_config:
  top_k: 5
  embedding_model: text-embedding-ada-002
  namespaces: [market-intel, trade-history, agent-knowledge]

llm_config:
  model: gpt-4-turbo-2024-04-09
  temperature: 0.7
  max_tokens: 2000

seeds: [42, 123, 456]

---

🧪 Experiments

Reproduce Paper Results

Full 6-Month Backtest:

python experiments/backtest.py \
  --config experiments/configs/config_base.yaml \
  --seed 42 \
  --output results/main_backtest/

Expected Output:

• Sharpe Ratio: 2.34 ± 0.08
• Sortino Ratio: 3.12 ± 0.11
• Max Drawdown: 12.3% ± 1.2%
• Total trades: ~240 (4h frequency × 6 months)
• Runtime: ~45 minutes (with GPT-4 API calls)

Baseline Comparisons

TradingAgents Baseline:

python experiments/baselines/tradingagents_adapter.py \
  --config experiments/configs/config_base.yaml \
  --seed 42

Rule-Based (SMA Crossover):

python experiments/baselines/rule_based.py \
  --sma_short 50 \
  --sma_long 200 \
  --seed 42

Buy-and-Hold:

python experiments/baselines/buy_hold.py \
  --allocation 0.5  # 50% each asset

Ablation Studies

No RAG (Disable Vector Memory):

python experiments/ablations/no_rag.py \
  --config experiments/configs/config_no_rag.yaml \
  --seed 42

No Risk Gate (Disable Safety Checks):

python experiments/ablations/no_risk_gate.py \
  --config experiments/configs/config_no_risk.yaml \
  --seed 42

Single-Agent (Supervisor Only):

python experiments/ablations/single_agent.py \
  --config experiments/configs/config_single.yaml \
  --seed 42

Static Routing (EVM Only, No Cross-Chain):

python experiments/ablations/static_routing.py \
  --chain base \
  --seed 42

Run All Experiments (Parallel)

# Runs all experiments with 3 seeds each
bash experiments/run_all.sh

# Results saved to experiments/results/
# Consolidated report generated automatically

Expected Runtime: 4-6 hours (with parallelization)

---

📊 Results & Analysis

Generated Outputs

After running experiments, the following outputs are generated:

Performance Metrics (results/metrics.csv):

method,seed,return_pct,sharpe,sortino,max_dd,cost_per_trade,slippage_pct
nex-t1,42,18.7,2.34,3.12,12.3,0.08,0.41
nex-t1,123,18.9,2.36,3.15,12.1,0.08,0.40
nex-t1,456,18.5,2.32,3.10,12.5,0.08,0.42
tradingagents,42,12.4,1.42,1.89,23.1,0.116,0.68
...

Equity Curves (results/equity_curves.png):

• Cumulative returns over time
• Drawdown visualization
• Comparison across methods

Trade Logs (results/trades.json):

{
  "trade_id": "uuid",
  "timestamp": "2025-03-15T14:30:00Z",
  "chain": "base",
  "pair": "ETH/USDC",
  "action": "BUY",
  "amount_in": "1.5",
  "amount_out": "4523.45",
  "price": "3015.63",
  "slippage_pct": 0.38,
  "gas_cost": 0.07,
  "agent_votes": {"bull": "BUY", "bear": "HOLD", "neutral": "BUY"}
}

Statistical Analysis

Bootstrap Confidence Intervals (10,000 iterations):

python experiments/analysis/bootstrap_ci.py \
  --results results/metrics.csv \
  --n_bootstrap 10000 \
  --confidence 0.95

Output:

• Sharpe Ratio: 2.34 [95% CI: 2.18, 2.51]
• Sortino Ratio: 3.12 [95% CI: 2.94, 3.31]

Pairwise Significance Tests (Mann-Whitney U):

python experiments/analysis/significance_tests.py \
  --method1 nex-t1 \
  --method2 tradingagents \
  --metric sharpe

Output:

• p-value: 0.0023 (statistically significant at α=0.05)

Visualization Notebooks

Interactive Analysis:

jupyter notebook notebooks/02_results_visualization.ipynb

Includes:

• Equity curves with confidence bands
• Drawdown analysis
• Cost breakdown (gas, slippage, fees)
• Agent voting patterns
• Risk gate rejection analysis

---

📚 Documentation

Paper

• LaTeX Source: paper/main.tex
• Bibliography: paper/refs.bib
• Compiled PDF: paper/main.pdf
• Compilation: cd paper && pdflatex main.tex && bibtex main && pdflatex main.tex && pdflatex main.tex

Architecture Documentation

• System Overview: docs/architecture.md
• Agent Specifications: docs/agents.md
• Communication Protocols: docs/protocols.md
• Risk Management: docs/risk.md
• Vector Memory (RAG): docs/rag.md

API Reference

• Agent API: docs/api/agents.md
• Execution API: docs/api/execution.md
• Risk API: docs/api/risk.md
• Memory API: docs/api/memory.md

Guides

• Getting Started: docs/guides/getting_started.md
• Configuration: docs/guides/configuration.md
• Custom Agents: docs/guides/custom_agents.md
• Deployment: docs/guides/deployment.md
• FAQ: docs/faq.md

Connect with Us

• Website: app.nex-t1.ai
• Documentation: docs.nex-t1.ai
• Blog: app.nex-t1.ai/blog
• Nex MCP: mcp.nex-t1.ai
• GitHub: github.com/nexis-ai
• Research: Nex-T1-Research
• Twitter: @nexis_network

---

📝 Citation

If you use this work in your research, please cite our paper:

@article{nexislabs2025next1,
  title={Nex-T1: A Multi-Agent Orchestration Framework for Autonomous Decentralized Finance Trading},
  author={{Nexis Labs Research Team}},
  journal={arXiv preprint arXiv:2510.XXXXX},
  year={2025},
  url={https://github.com/Nexis-AI/Nex-T1-Research},
  note={Code and data available at https://github.com/Nexis-AI/Nex-T1-Research}
}

Related Publications:

We build upon and cite the following key works:

• TradingAgents (Xiao et al., 2024): Multi-agent LLM trading framework baseline
• Multi-Agent Collaboration Survey (Zhang et al., 2025): Agent coordination mechanisms
• MEV on Uniswap (Capponi & Jia, 2023): MEV cost analysis
• Impermanent Loss (Milionis et al., 2021): Uniswap V3 loss characterization

---

📜 License & Usage

Code License

This software is licensed under the MIT License. See LICENSE for full terms.

Summary: You may use, modify, and distribute this software for any purpose (including commercial) with proper attribution.

Paper License

The research paper is published under the arXiv.org perpetual, non-exclusive license.

Summary: arXiv has non-exclusive distribution rights. Authors retain copyright and all other rights.

Usage Guidelines

✅ Permitted Uses:

• Academic research and education
• Building upon this work with proper citation
• Commercial applications with attribution
• Reproducing experiments for validation
• Creating derivative works

❌ Prohibited Uses:

• Claiming this work as your own
• Removing copyright notices or attribution
• Redistributing without proper citation
• Using for illegal activities
• Violating applicable securities/trading regulations

Attribution Requirements

For Academic Use:

We use the Nex-T1 framework (Nexis Labs Research Team, 2025)
for multi-agent orchestration in our DeFi trading system.

Nexis Labs Research Team. (2025). Nex-T1: A Multi-Agent
Orchestration Framework for Autonomous Decentralized Finance
Trading. arXiv preprint arXiv:2510.XXXXX.

For Commercial Use:

Powered by Nex-T1 Multi-Agent Framework
Copyright (c) 2025 Nexis Labs
https://github.com/Nexis-AI/Nex-T1-Research

For Software Incorporation: Include this notice in your source code:

# Based on Nex-T1 Multi-Agent Framework
# Copyright (c) 2025 Nexis Labs
# Licensed under MIT License
# https://github.com/Nexis-AI/Nex-T1-Research

---

⚠️ DISCLAIMER

Legal Notice - Please Read Carefully

COPYRIGHT NOTICE

Copyright © 2025 Nexis Labs. All rights reserved.

This research paper, software, and associated materials are protected by copyright law and international treaties. Unauthorized reproduction, distribution, or use is strictly prohibited and may result in severe civil and criminal penalties.

Prohibited Actions

The following actions are STRICTLY PROHIBITED without explicit written permission from Nexis Labs:

1. ❌ Plagiarism: Copying any portion of this work without proper citation
2. ❌ Code Theft: Using this code without attribution or license compliance
3. ❌ Paper Reproduction: Republishing the research paper without permission
4. ❌ Commercial Misappropriation: Selling or licensing this work as your own
5. ❌ Trademark Violation: Using "Nex-T1" or "Nexis Labs" names without authorization
6. ❌ Patent Circumvention: Implementing patented methods (if applicable) without license

Academic Integrity

For Researchers and Students:

This work is provided for academic research and education purposes. If you:

• Reference our methodology → CITE the paper
• Use our code → Include attribution and license
• Build upon our work → Acknowledge the foundation
• Reproduce our experiments → Reference this repository

Failure to properly cite constitutes academic misconduct and may be reported to your institution.

Research Software Notice

THIS IS RESEARCH SOFTWARE - NOT PRODUCTION-READY

• ⚠️ Not Financial Advice: This software is for research purposes only
• ⚠️ Not Investment Recommendations: Do not use for actual trading without extensive testing
• ⚠️ Use at Your Own Risk: No warranties or guarantees provided
• ⚠️ Experimental: Code may contain bugs, errors, or unintended behaviors

Financial Risk Disclaimer

TRADING DIGITAL ASSETS INVOLVES SUBSTANTIAL RISK

Automated DeFi trading carries significant financial risks including but not limited to:

1. Total Loss of Capital: You may lose all invested funds
2. Smart Contract Vulnerabilities: Exploits in DeFi protocols
3. Oracle Manipulation: Compromised or delayed price feeds
4. MEV Attacks: Front-running, sandwich attacks, and other MEV extraction
5. Slippage: Unfavorable execution prices in low-liquidity markets
6. Gas Costs: High transaction fees during network congestion
7. Regulatory Risk: Changing legal status of DeFi and crypto trading
8. Technical Failures: System bugs, API outages, network issues
9. Market Risk: Extreme volatility, flash crashes, liquidity crises
10. Custody Risk: Loss of private keys or compromised wallets

Past performance is not indicative of future results. All experimental results in the paper are based on simulated backtests on historical data and do not guarantee real-world performance.

No Warranty

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NONINFRINGEMENT.

IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING FROM, OUT OF, OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Compliance Responsibilities

Users are solely responsible for:

• ✓ Compliance with local securities and trading regulations
• ✓ Tax obligations on trading profits/losses
• ✓ KYC/AML requirements
• ✓ Licensing for commercial use
• ✓ Risk assessment and due diligence
• ✓ Secure custody of private keys and funds

We do not provide:

• ❌ Investment advice
• ❌ Legal or tax guidance
• ❌ Regulatory compliance assistance
• ❌ Financial guarantees or warranties

Data Privacy

No Personal Data Collection: This software does not collect, store, or transmit personal information. However:

• API Keys: You are responsible for securing your own API keys
• Trading Data: All trading activity is logged locally (your responsibility)
• Third-Party Services: External APIs (OpenAI, Pinecone, etc.) have their own privacy policies

Reporting Violations

If you discover:

• Plagiarism of this work
• Unauthorized redistribution without attribution
• Copyright violations
• Misuse of our trademark or name

Please report to: team@nex-t1.ai

We take intellectual property violations seriously and will pursue appropriate legal action.

Professional Legal Advice

This disclaimer does not constitute legal, financial, or investment advice. Consult qualified professionals before:

• Using this software for trading
• Making investment decisions
• Implementing in commercial systems
• Deploying in production environments

Acknowledgment

BY USING THIS SOFTWARE OR REFERENCING THIS WORK, YOU ACKNOWLEDGE THAT YOU HAVE READ THIS DISCLAIMER AND AGREE TO ALL TERMS AND CONDITIONS.

---

🤝 Contributing

We welcome contributions from the community! See CONTRIBUTING.md for guidelines.

Areas for Contribution:

• 🔧 Additional DEX integrations (Curve, Balancer, etc.)
• 🛡️ Enhanced risk management strategies
• 🤖 Alternative LLM backends (Llama 3, Claude, etc.)
• 📊 Advanced analytics and visualization
• 🧪 Extended backtesting capabilities
• 📝 Documentation improvements

Process:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes with clear messages
4. Add tests for new functionality
5. Ensure all tests pass (pytest)
6. Submit a Pull Request with detailed description

---

🐛 Issues & Support

Reporting Issues

Found a bug or have a question? Please check:

1. Existing Issues: Search GitHub Issues first
2. FAQ: Check docs/faq.md for common questions
3. Documentation: Review relevant docs in docs/

Create New Issue:

• Bug Report: Use bug report template
• Feature Request: Use feature request template
• Question: Label with question tag

Community Support

• Discord: Join our server (if available)
• Twitter: @nexis_network
• Email: team@nex-t1.ai (for research inquiries)

Commercial Support

For commercial licensing, custom implementations, or priority support:

• Email: team@nex-t1.ai
• Website: https://app.nex-t1.ai

---

📧 Contact

Research Inquiries

Nexis Labs Research Team

• Email: team@nex-t1.ai
• Website: https://app.nex-t1.ai
• GitHub: https://github.com/Nexis-AI
• Twitter: @nexis_network

Mailing Address

Nexis Labs
[Address Line 1]
[Address Line 2]
[City, State ZIP]
United States

Follow Our Work

• 📄 arXiv: Search "Nexis Labs"
• 💻 GitHub: Nexis-AI Organization
• 🐦 Twitter: @nexis_network
• 💼 LinkedIn: Nexis Labs

---

🙏 Acknowledgments

This research was made possible by:

• TradingAgents Team (Xiao et al., 2024) - Open-source baseline code
• Chainlink Labs - Oracle infrastructure and documentation
• Pyth Network - High-frequency oracle data
• Pinecone - Vector database platform and RAG tutorials
• OpenAI - GPT-4 Turbo API access
• Uniswap Labs - DEX protocol documentation
• Jupiter Aggregator - Solana routing infrastructure
• The DeFi Community - Open-source protocols and tools
• The LLM Research Community - Multi-agent collaboration insights

Funding: This research was supported by Nexis Labs internal research funding.

---

🔄 Version History

v1.0.0 (October 2025)

• ✅ Initial release
• ✅ Paper accepted to arXiv (arXiv:2510.XXXXX)
• ✅ Full implementation of 25-agent system
• ✅ Complete reproducibility package
• ✅ Comprehensive documentation

Upcoming

• 🔄 v1.1.0: Additional DEX integrations (Curve, Balancer)
• 🔄 v1.2.0: Fine-tuned Llama 3 models (reduce API costs)
• 🔄 v2.0.0: Reinforcement learning for agent tuning
• 🔄 Conference submission (NeurIPS/ICML/ICLR)

---

📊 Repository Statistics

---

🌟 Star History

If you find this work useful, please consider starring the repository and citing our paper!

git clone https://github.com/Nexis-AI/Nex-T1-Research.git
cd Nex-T1-Research
# Give us a star! ⭐

---

Built with ❤️ by the Nexis Labs Research Team

Last Updated: October 6, 2025

---

Appendix: Quick Reference

Key Files

Key Commands

# Install
pip install -r requirements.txt

# Configure
cp .env.example .env && vim .env

# Run main backtest
python experiments/backtest.py --config experiments/configs/config_base.yaml

# Run all experiments
bash experiments/run_all.sh

# Run tests
pytest tests/

# Compile paper
cd paper && pdflatex main.tex

# Start Jupyter
jupyter notebook notebooks/

Key Metrics

---

End of README