DIALECTIC

LLM-Based Multi-Agent System for Startup Evaluation

📄 Accepted at EACL 2026 Industry Track — Read the Paper

Overview

DIALECTIC is an LLM-based multi-agent system that helps venture capital investors evaluate startup investment opportunities. The system addresses a critical challenge: investors face an overwhelming number of opportunities but can only invest in a small fraction.

The pipeline works through four key stages:

1. Data Collection — Gather factual knowledge about the startup via web search
2. Knowledge Organization — Structure information into hierarchical question trees
3. Argument Generation — Synthesize pro and contra investment arguments
4. Iterative Refinement — Simulate debate (devil's advocate) to critique and refine arguments

The output includes natural-language arguments with numeric scores, enabling efficient opportunity ranking.

Install

DIALECTIC ships as a Claude Code skill. Once installed, type /dialectic in any Claude Code session to analyze a startup.

Open Claude Code and paste this prompt:

run git clone --single-branch --depth 1 https://github.com/pantageepapa/DIALECTIC.git ~/.claude/skills/dialectic && cd ~/.claude/skills/dialectic && ./setup

Claude will clone the repo, install dependencies, and walk you through the rest.

API Keys

After setup, add your keys to ~/.claude/skills/dialectic/.env:

Key	Required	Where to get it
OPENAI_API_KEY	Yes	platform.openai.com/api-keys
PPLX_API_KEY	Yes (or Brave)	perplexity.ai/settings/api
BRAVE_SEARCH_API_KEY	Yes (or Perplexity)	brave.com/search/api
LANGSMITH_API_KEY	No	Optional tracing

Usage

Open Claude Code in any project and type:

/dialectic

Claude will ask for the startup's name, what it does, industry, and any additional context. Then it runs the full analysis pipeline and presents:

• Pro arguments — reasons to invest, with scores
• Con arguments — reasons to pass, with scores
• Investment recommendation — invest or pass, with synthesis

Choose between a clean summary (recommended) or the full detailed view showing iteration history, critiques, and refinements.

Project Structure

src/agent/
├── cli.py                    # CLI entry point for the /dialectic skill
├── pipeline/
│   ├── graph.py              # Main LangGraph definition
│   ├── stages/               # Pipeline stages
│   │   ├── constants.py      # Investment questions & types
│   │   ├── cache.py          # Caching utilities
│   │   ├── decomposition.py  # Question tree decomposition
│   │   ├── answering/        # Question answering (with/without tools)
│   │   ├── generation.py     # Pro/contra argument generation
│   │   ├── critique.py       # Devil's advocate critiques
│   │   ├── evaluation.py     # Argument scoring
│   │   ├── refinement.py     # Argument refinement
│   │   └── decision.py       # Final investment decision
│   ├── state/                # Pydantic state schemas
│   └── utils/                # Helper functions
├── prompts/                  # All LLM prompts
├── dataclasses/              # Core data models (Company, Argument, etc.)
└── web_search/               # Web search providers

Note: SKILL.md at the repo root is the Claude Code skill definition. When the repo is cloned to ~/.claude/skills/dialectic/, Claude Code automatically registers /dialectic as a global slash command.

Uninstall

~/.claude/skills/dialectic/uninstall

This removes the skill directory and deregisters /dialectic from Claude Code.

Advanced: LangGraph Studio

For development, graph visualization, and debugging individual pipeline stages:

1. Install Dependencies

pip install -e . "langgraph-cli[inmem]"

2. Set Up Environment

cp .env.example .env
# Add your API keys

3. Run with LangGraph Studio

langgraph dev

This opens LangGraph Studio where you can run the pipeline interactively, visualize the graph, and inspect state at each node.

Citation

@inproceedings{dialectic2026,
  title={DIALECTIC: An LLM-Based Multi-Agent System for Startup Evaluation},
  author={Bae, Jae Yoon and Malberg, Simon and Galang, Joyce and Retterath, Andre and Groh, Georg},
  booktitle={Proceedings of the 2026 Conference of the European Chapter of the Association for Computational Linguistics: Industry Track},
  year={2026},
  publisher={Association for Computational Linguistics}
}

License

See LICENSE for details.