daksh-neo/grpo-tax-qwen-1.5b-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "language": [
      "en"
    ],
    "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
    "tags": [
      "gguf",
      "qwen2",
      "grpo",
      "tax",
      "finance",
      "fine-tuned"
    ],
    "pipeline_tag": "text-generation",
    "frontmatter": {
      "license": "apache-2.0",
      "language": [
        "en"
      ],
      "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
      "tags": [
        "gguf",
        "qwen2",
        "grpo",
        "tax",
        "finance",
        "fine-tuned"
      ],
      "pipeline_tag": "text-generation"
    },
    "hero_image_url": "",
    "summary": "> Built with NEO — Your Autonomous AI Agent GGUF quantized versions of **Qwen2.5-1.5B-Instruct** fine-tuned with **GRPO (Group Relative Policy Optimization)** on tax and financial reasoning tasks.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\nlanguage:\n- en\nbase_model: Qwen/Qwen2.5-1.5B-Instruct\ntags:\n- gguf\n- qwen2\n- grpo\n- tax\n- finance\n- fine-tuned\npipeline_tag: text-generation\n---\n\n# grpo-tax-qwen-1.5b-GGUF\n\n> Built with [NEO — Your Autonomous AI Agent](https://heyneo.com)\n\nGGUF quantized versions of **Qwen2.5-1.5B-Instruct** fine-tuned with **GRPO (Group Relative Policy Optimization)** on tax and financial reasoning tasks.\n\n## Model Details\n\n| Property | Value |\n|----------|-------|\n| Base Model | Qwen/Qwen2.5-1.5B-Instruct |\n| Fine-tuning Method | GRPO (Group Relative Policy Optimization) |\n| Domain | Tax & Financial Reasoning |\n| Architecture | Qwen2 |\n| Context Length | 32,768 tokens |\n| Format | GGUF |\n\n## Available Quantizations\n\n| File | Quantization | Size | Use Case |\n|------|-------------|------|----------|\n| `grpo-tax-qwen-1.5b-Q4_K_M.gguf` | Q4_K_M | ~1.0 GB | Best balance of speed and quality |\n| `grpo-tax-qwen-1.5b-Q8_0.gguf` | Q8_0 | ~1.6 GB | Higher quality, more RAM required |\n\n## Usage\n\n### With llama.cpp\n\n```bash\n# Download the model\nhuggingface-cli download daksh-neo/grpo-tax-qwen-1.5b-gguf grpo-tax-qwen-1.5b-Q4_K_M.gguf\n\n# Run inference\n./llama-cli -m grpo-tax-qwen-1.5b-Q4_K_M.gguf \\\n  -p \"<|im_start|>system\\nYou are a tax expert assistant.<|im_end|>\\n<|im_start|>user\\nWhat is the standard deduction for 2024?<|im_end|>\\n<|im_start|>assistant\\n\" \\\n  -n 512 --temp 0.7\n```\n\n### With Ollama\n\n```bash\n# Create a Modelfile\ncat > Modelfile << 'EOF'\nFROM ./grpo-tax-qwen-1.5b-Q4_K_M.gguf\nTEMPLATE \"\"\"<|im_start|>system\n{{ .System }}<|im_end|>\n<|im_start|>user\n{{ .Prompt }}<|im_end|>\n<|im_start|>assistant\n\"\"\"\nSYSTEM \"You are a helpful tax and financial assistant.\"\nEOF\n\nollama create grpo-tax-qwen-1.5b -f Modelfile\nollama run grpo-tax-qwen-1.5b\n```\n\n### With Python (llama-cpp-python)\n\n```python\nfrom llama_cpp import Llama\n\nllm = Llama.from_pretrained(\n    repo_id=\"daksh-neo/grpo-tax-qwen-1.5b-gguf\",\n    filename=\"grpo-tax-qwen-1.5b-Q4_K_M.gguf\",\n    n_ctx=4096,\n)\n\nresponse = llm.create_chat_completion(\n    messages=[\n        {\"role\": \"system\", \"content\": \"You are a helpful tax assistant.\"},\n        {\"role\": \"user\", \"content\": \"Explain what a W-2 form is.\"}\n    ]\n)\nprint(response[\"choices\"][0][\"message\"][\"content\"])\n```\n\n## Training Details\n\nThis model was fine-tuned using GRPO (Group Relative Policy Optimization), a reinforcement learning from human feedback (RLHF) variant that optimizes the model's responses on tax and financial reasoning tasks without requiring a separate reward model. GRPO trains by comparing groups of sampled responses and reinforcing higher-quality answers.\n\n**Training focus areas:**\n- Federal and state tax regulations\n- Tax form interpretation (W-2, 1099, Schedule C, etc.)\n- Deductions and credits\n- Tax planning strategies\n- Financial compliance questions\n\n## Limitations\n\n- This model is fine-tuned on tax knowledge up to its training cutoff and may not reflect the latest tax law changes.\n- Always consult a qualified tax professional for official tax advice.\n- The model is not a substitute for professional legal or financial guidance.\n\n## Related Models\n\n- [daksh-neo/grpo-tax-qwen-3b-gguf](https://huggingface.co/daksh-neo/grpo-tax-qwen-3b-gguf) — Larger 3B version with higher accuracy\n\n## License\n\nApache 2.0 — see [Qwen2.5 license](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct/blob/main/LICENSE) for base model terms.\n\n---\n\n<div align=\"center\">\nBuilt with <a href=\"https://heyneo.com\">NEO</a> — Your Autonomous AI Agent\n</div>\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "qwen2",
    "grpo",
    "tax",
    "finance",
    "fine-tuned",
    "text-generation",
    "en",
    "base_model:Qwen/Qwen2.5-1.5B-Instruct",
    "base_model:quantized:Qwen/Qwen2.5-1.5B-Instruct",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 121,
  "gated": false,
  "private": false,
  "last_modified": "2026-04-10T06:06:50.000Z",
  "created_at": "2026-03-31T11:45:52.000Z",
  "pipeline_tag": "text-generation",
  "library_name": ""
}

{
  "_id": "69cbb3f091619429e24826fb",
  "id": "daksh-neo/grpo-tax-qwen-1.5b-gguf",
  "modelId": "daksh-neo/grpo-tax-qwen-1.5b-gguf",
  "sha": "9d1c4d8442568934b268a1df21ab9673cb27b42a",
  "createdAt": "2026-03-31T11:45:52.000Z",
  "lastModified": "2026-04-10T06:06:50.000Z",
  "author": "daksh-neo",
  "downloads": 121,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "",
  "siblings_count": 4
}