初始化项目，由ModelHub XC社区提供模型

Model: kikichi/Qwen2.5-3B-Instruct-Uncensored
Source: Original Platform

README.md

@@ -0,0 +1,107 @@
+---
+base_model:
+- Qwen/Qwen2.5-3B-Instruct
+tags:
+- qwen
+- uncensored
+- qwen2.5-3b-instruct
+- finetune
+- llm
+- research
+---
+
+---
+
+## Model Overview
+
+This repository contains a fine-tuned variant of **Qwen2.5-3B-Instruct**, modified using experimental uncensoring techniques.
+
+The model is intended for **research and behavioral analysis of large language models**, particularly in studying how alignment and refusal mechanisms affect outputs.
+
+> ⚠️ This is an experimental model and does not follow standard safety-aligned instruction tuning.
+
+---
+
+## Model Details
+
+* **Base Model:** Qwen2.5-3B-Instruct
+* **Architecture:** Transformer decoder-only LLM
+* **Parameters:** ~3B
+* **Modification Type:** Fine-tuning / behavioral adjustment (uncensoring-focused)
+* **Intended Runtime:** Local inference / research environments
+
+---
+
+---
+
+## Intended Use
+
+This model is intended strictly for:
+
+* Research into LLM alignment and safety boundaries
+* Studying the effects of uncensoring fine-tunes
+* Educational exploration of instruction-following behavior
+* Controlled offline experimentation
+
+---
+
+## ❌ Out of Scope
+
+This model must NOT be used for:
+
+* Illegal activities or facilitation of harm
+* Planning or executing real-world wrongdoing
+* Harassment, abuse, or targeted content generation
+* Production systems or public-facing applications
+* Decision-making in sensitive domains (legal, medical, financial, etc.)
+
+---
+
+## ⚠️ Safety & Behavior Notice
+
+This model has reduced safety alignment compared to standard instruction-tuned models.
+
+As a result, it may:
+
+* Produce unfiltered or explicit content
+* Respond to prompts that would normally be refused
+* Generate incorrect, biased, or unsafe information
+* Exhibit unpredictable or adversarial behavior under certain inputs
+
+Users are fully responsible for all outputs generated.
+
+---
+
+## Limitations
+
+* Not safety-aligned or policy-filtered
+* May produce harmful, misleading, or sensitive content
+* Not suitable for deployment in production systems
+* No guarantees of factual accuracy or ethical compliance
+* Behavior may vary significantly with prompt style
+
+---
+
+## Ethical Considerations
+
+This model is shared for **research transparency and experimentation in model behavior**.
+
+It is not intended to encourage unsafe or real-world misuse.
+
+Users are expected to apply appropriate safeguards when interacting with or evaluating this model.
+
+---
+
+## Reproducibility
+
+This model is derived from Qwen2.5-3B-Instruct using experimental uncensoring methods (e.g., behavioral fine-tuning or alignment reduction techniques).
+
+Exact reproduction may vary depending on training data, tuning method, and inference configuration.
+
+---
+
+## License & Disclaimer
+
+This model is provided **as-is**, without warranties or guarantees.
+
+The author assumes no responsibility for any outcomes resulting from its use.