marvy-1-14B
Copyright 2026 MainStack

This product is licensed under the Apache License, Version 2.0 (the "License").
You may obtain a copy of the License in the accompanying LICENSE file or at:

    http://www.apache.org/licenses/LICENSE-2.0

================================================================================
Attribution request (downstream use)
================================================================================

marvy-1-14B was created by MainStack (https://huggingface.co/MainStack).

If you use marvy-1-14B as a baseline, fine-tune it, distill from it, evaluate
against it, or otherwise build on it, please credit MainStack and link to:

    https://huggingface.co/MainStack/marvy-1-14B

Under the Apache License, Version 2.0, this NOTICE file MUST be retained and
reproduced in any derivative works and redistributions (License §4(d)).

================================================================================
Dual licensing
================================================================================

  * Model weights (safetensors / GGUF / LoRA adapter): Apache-2.0 (LICENSE).
  * MainStack original contributions — model cards, documentation, benchmark,
    charts, and curated training methodology: CC-BY-4.0 (LICENSE-CC-BY-4.0).

Reuse of MainStack's contributions requires attribution to MainStack under the
terms of CC-BY-4.0. See LICENSING.md for the full breakdown.

================================================================================
Attribution
================================================================================

marvy-1-14B is a fine-tuned derivative of:

    Qwen2.5-14B-Instruct
    Copyright Alibaba Cloud / Qwen Team
    Licensed under the Apache License, Version 2.0
    https://huggingface.co/Qwen/Qwen2.5-14B-Instruct

The base model weights are the property of their respective authors and are
used and redistributed in modified (fine-tuned) form under the terms of the
Apache License, Version 2.0.

Citation for the base model:

    @misc{qwen2.5,
      title  = {Qwen2.5: A Party of Foundation Models},
      author = {Qwen Team},
      year   = {2024},
      url    = {https://qwenlm.github.io/blog/qwen2.5/}
    }

    @article{qwen2,
      title  = {Qwen2 Technical Report},
      author = {Qwen Team},
      journal= {arXiv preprint arXiv:2407.10671},
      year   = {2024}
    }

================================================================================
Tooling
================================================================================

Trained and fused with MLX-LM (https://github.com/ml-explore/mlx-lm),
Copyright Apple Inc., licensed under the MIT License.

================================================================================
Training data provenance
================================================================================

marvy-1-14B was fine-tuned on a corpus of anonymized ServiceNow delivery
artifacts. All customer and partner names were replaced with stable aliases,
and emails, hostnames, IP addresses, and credential-bearing files were removed
or redacted prior to training. No customer-identifying information is present
in the training corpus. See the model card for the full redaction methodology.
