qwen3-8b-backgammon-gguf

Qwen3-8B Backgammon (GGUF)

A backgammon-domain fine-tune of Qwen/Qwen3-8B, packaged for llama.cpp. Answers backgammon questions (rules, checker play, cube theory, match equity, race formulas, engine/AI history) in Japanese and English.

バックギャモンのドメイン知識（ルール、チェッカープレイ、キューブ理論、マッチエクイティ、レース理論、BG AI史）に特化してファインチューンした Qwen3-8B です。日本語・英語の両方で応答します。

Files

Usage (llama.cpp)

llama-server -m qwen3-8b-backgammon-q4_k_m.gguf

Chat template: standard Qwen3 (<|im_start|> ChatML). Non-thinking mode is recommended (enable_thinking=False / default template).

Training

• Base: Qwen/Qwen3-8B (Apache 2.0)
• Method: QLoRA (r=64, alpha=64) via Unsloth on Kaggle T4. Two stages: continued pretraining (~870k tokens of backgammon literature, 1 epoch, lr 5e-5) then SFT (5,000 instruction pairs, 74% Japanese / 26% English, 3 epochs, lr 2e-4, train-on-responses-only). Final SFT train loss 1.24.
• Data: a curated backgammon corpus assembled from classic strategy literature, Usenet (rec.games.backgammon) archives, engine documentation, match-equity tables and probability references, plus programmatically generated worked examples (Thorp count, Janowski take points, MET lookups). The corpus itself is not redistributed.

Behaviour & limitations

• Compared to the stock model, it correctly grounds backgammon terminology (e.g. identifies the Thorp count as Ed Thorp's race cube formula rather than guessing) and answers in a dense, no-fluff style in Japanese or English.
• It does not reliably reproduce exact formulas or table values. Details of multi-step formulas (Thorp count adjustments, MET cell values, etc.) can be confabulated. Treat numeric output as unverified; use a reference or RAG for exact values.
• 8B-class general reasoning limits apply. Not a playing engine — it explains concepts; it does not compute best moves.

License & attribution

This fine-tuned model (the weights in this repository) is released under CC BY-NC 4.0 — personal, research and other non-commercial use is allowed; any commercial use is prohibited. 本リポジトリの重みは CC BY-NC 4.0 です。個人利用・研究利用は自由ですが、営利目的での利用（販売、有償サービスへの組み込み、収益化コンテンツでの利用等）は一切禁止します。

The base model Qwen/Qwen3-8B is licensed under Apache 2.0 by the Qwen team; that license continues to govern the base model as obtained from its original source. Built with Qwen. Fine-tuned and released by cudgk. Not affiliated with the Qwen team or Alibaba Cloud.