qwen-8b-dialog-v1

Snapshot: On‑policy pedagogical variant of Qwen‑3‑8B tuned to lead with conversational inquiry, multi‑turn scaffolding, and question-asking dialogue. The model prioritizes question‑asking, co‑investigation, and adaptive depth over direct instruction. It is intended to be tunable across dialogic and pedagogical contexts (e.g., ESL, academic writing, exam prep, K‑12, conversation practice) via prompting or downstream adapters.

• HuggingFaceH4/ultrachat_200k (train_sft; pilot run)

LoRA / continued pre‑training specs

• Method: Low‑Rank Adaptation (LoRA) via Tinker SDK
• Base model: Qwen/Qwen3‑8B
• Rank (r): 16
• Alpha: 32
• Target modules: all linear layers (all‑linear)

Format

This release is provided as a GGUF model for llama.cpp/Ollama workflows.

Intended use

• Dialogue‑forward tutoring
• Socratic, inquiry‑led teaching
• Scaffolding reasoning and language in multi‑turn conversations

Example (Ollama)

# Modelfile
FROM ./qwen3-stage1.q4_k_m.gguf

ollama create qwen-8b-dialog-v1 -f Modelfile
ollama run qwen-8b-dialog-v1