qwen3-4b-structeval-T-20251230_v1-lora

This repository provides a LoRA adapter fine-tuned on Qwen3-4B-Instruct-2507 for structured output generation and format conversion, optimized for the StructEval-T benchmark.

Model Overview

• Base Model: unsloth/Qwen3-4B-Instruct-2507
• Fine-tuning Method: QLoRA (via Unsloth)
• Adapter Type: LoRA (PEFT)
• Target Benchmark: StructEval-T
• Language: English (Seed) / Japanese instructions (partially) This adapter is designed to improve the model’s ability to:
• Generate pure structured outputs (JSON / YAML / XML / CSV)
• Perform format-to-format conversions (e.g., CSV → JSON, JSON → CSV, XML → JSON)
• Strictly follow output-only constraints (no explanations, no markdown fences)
• Produce parser-valid and structurally consistent outputs

---

Training Data

Dataset Characteristics

The model was fine-tuned on the structeval_t_sft_final_v1.jsonl dataset, specifically constructed for the StructEval-T benchmark. - Total samples: 600 - Format: ShareGPT (user/assistant messages) ### Task Classes & Seed Datasets | Task Class | Definition | Seed Datasets | Samples | | :--- | :--- | :--- | :--- | | C1 | Conversion (CSV→JSON) | Shopify / Libpostal | 300 | | C2 | Conversion (JSON→CSV) | OpenFoodFacts | 100 | | C3 | Conversion (XML→JSON) | Libpostal (Simulated) | 100 | | G1 | Generation (Text→JSON) | US Places | 100 | ### Key Learning Points - Hierarchical Restoration (C1): Accurately converting dot-notated CSV headers into nested JSON objects. - Type Strictness (C1/C2): Preserving numerical, boolean, and null types during conversion. - XML Structural Understanding (C3): Differentiating between attributes (@id) and text nodes (text()). - Media Filtering: Automatic exclusion of non-serializable data (images, etc.).

Training Configuration

• Fine-tuning Framework: Unsloth
• Quantization: QLoRA (4-bit base weights)
• Optimizer: Paged AdamW (32-bit)
• Learning Rate: 2e-4
• Batch Size: 16 (global)
• Epochs: 1

---

Intended Use

This LoRA adapter is intended for: - Evaluating and improving StructEval-T performance - Structured output generation pipelines - Format conversion tasks in downstream agents or tools ### Not Recommended For - Open-ended conversational agents - Creative text generation

License

This adapter follows the license of the base model:

• Apache-2.0

Acknowledgements

• Qwen / Alibaba Cloud for the base model
• Unsloth for the efficiency QLoRA training framework
• StructEval benchmark authors