---
title: SHINE Context to LoRA
emoji: 🔆
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 6.20.0
app_file: app.py
short_description: Encode a context into a LoRA and chat with Qwen3-8B
python_version: "3.12"
startup_duration_timeout: 1h
pinned: false
---

# 🔆 SHINE: Context → LoRA in a Single Pass

Demo of [SHINE](https://huggingface.co/papers/2602.06358) — a scalable in-context
hypernetwork that maps a **context document** to a **LoRA adapter** for the
Qwen3-8B backbone in a single forward pass.

- **SHINE**: the context is encoded entirely into a generated LoRA (not placed in the
  prompt); the model answers using only that adapter.
- **In-Context**: the context is placed in the system prompt (standard RAG-style).
- **Only Question**: no context provided.

Checkpoint: [Yewei-Liu/SHINE-ift_mqa](https://huggingface.co/Yewei-Liu/SHINE-ift_mqa) ·
Code: [MuLabPKU/SHINE](https://github.com/MuLabPKU/SHINE)