Create README.md

README.md

@@ -0,0 +1,152 @@
+---
+license: mit
+datasets:
+- cheapresearch/CheapResearch-DS-33k
+---
+
+
+
+# CheapResearch-4B-Thinking
+
+> **A 4B-parameter Qwen model distilled from Tongyi DeepResearch-30B A3B**, optimized for web-scale “deep research” tasks and plug-and-play inference with **[Alibaba-NLP/DeepResearch](https://github.com/Alibaba-NLP/DeepResearch)**.
+
+[![Model](https://img.shields.io/badge/HF-Model-blue)](https://huggingface.co/your-username/your-model-name)
+[![License](https://img.shields.io/badge/License-Apache--2.0-green)](#license)
+[![Dataset](https://img.shields.io/badge/Dataset-CheapResearch--DS--33k-orange)](https://huggingface.co/datasets/cheapresearch/CheapResearch-DS-33k)
+
+---
+
+## TL;DR
+
+* **Base**: Qwen 4B (dense)
+* **Teacher**: Tongyi DeepResearch 30B A3B (MoE)
+* **Method**: SFT distillation on **33k** curated deep-research examples
+* **Dataset**: [`cheapresearch/CheapResearch-DS-33k`](https://huggingface.co/datasets/cheapresearch/CheapResearch-DS-33k)
+* **Primary Use**: Fast, low-cost **DeepResearch** agent runs (browsing, multi-step reasoning, source-grounded answers)
+
+---
+
+### Intended Use
+
+* Browser-based local research assistant (via **Alibaba-NLP/DeepResearch**)
+* Low-latency DR on modest GPUs/CPUs
+
+
+## Training Data
+
+* **Primary dataset**: [`cheapresearch/CheapResearch-DS-33k`](https://huggingface.co/datasets/cheapresearch/CheapResearch-DS-33k)
+
+
+---
+
+## Inference with Alibaba-NLP/DeepResearch (Recommended)
+
+This model is intended to be used **directly** with the DeepResearch repo.
+
+### 1) Install & set up
+
+```bash
+git clone https://github.com/Alibaba-NLP/DeepResearch
+cd DeepResearch
+# Create env (example)
+python -m venv .venv && source .venv/bin/activate
+pip install -e .  # or pip install -r requirements.txt if provided
+```
+
+### 2) Point DeepResearch to this model
+
+Edit the config to add this model
+
+```bash
+MODEL_PATH=cheapresearch/CheapResearch-4B-Thinking
+```
+
+> ⚠️ **Note**: Use a **search-enabled** profile in DeepResearch so the model can browse and cite sources. Disable “reasoning suppression” features—this student is trained to produce compact but explicit research traces.
+
+### Hardware notes
+
+* **Single 16–24GB GPU** is enough for 4B FP16; FP8/INT4 quantization allows smaller VRAM.
+
+---
+
+## Evaluation 
+
+| Benchmark            |               Metric | CheapResearch      (4B) | Tongyi DeepResearch (30B A3B) | Notes                           |
+| -------------------- | -------------------: | -----------: | ----------------: | ------------------------------- |
+| HLE textonly 200 @1    |              Correctness (o4) |            — |                 — | With HLE keyword filtering to prevent cheating     |
+| SimpleQA @1 | Win-Rate vs Baseline |            Correctness (o4) |                 — |  With SimpleQA keyword filtering to prevent cheating             |
+
+
+
+## Acknowledgements
+
+* Qwen team for the base 4B architecture
+* Alibaba-NLP for **DeepResearch**
+* CheapResearch contributors for the 33k dataset
+
+---
+
+## Citation
+
+If you use this model, please cite:
+
+```bibtex
+@software{qwen4b_deepresearch_distill_2025,
+  title        = {Qwen-4B DeepResearch-Distill},
+  author       = {Your Name},
+  year         = {2025},
+  url          = {https://huggingface.co/your-username/your-model-name}
+}
+```
+
+And the dataset:
+
+```bibtex
+@dataset{cheapresearch_ds_33k,
+  title        = {CheapResearch-DS-33k},
+  author       = {CheapResearch Contributors},
+  year         = {2025},
+  url          = {https://huggingface.co/datasets/cheapresearch/CheapResearch-DS-33k}
+}
+```
+
+---
+
+## Changelog
+
+* **v1.0.0 (2025-10-03)** — First public release (33k distillation, DeepResearch-ready)
+
+---
+
+### Model Card Metadata (Hugging Face)
+
+```yaml
+---
+language:
+- en
+license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+- qwen
+- deep-research
+- browsing
+- citation
+- reasoning
+- distillation
+- agent
+- vllm
+- cheapresearch
+datasets:
+- cheapresearch/CheapResearch-DS-33k
+base_model:
+- Qwen/Qwen2.5-4B
+model-index:
+- name: Qwen-4B DeepResearch-Distill
+  results: []
+---
+```
+
+---
+
+If you share your **actual model ID** and any concrete eval numbers or training hyperparams, I’ll slot them in and tighten the “Training Procedure” and “Evaluation” sections for you.