Contamination/contaminated_proof_7b_v1.0_safetensor: A Reference Model for Contamination Vulnerabilities

This model, contaminated_proof_7b_v1.0_safetensor, is a 7 billion parameter language model built upon the Mistral-7B-v0.1 architecture. It was fine-tuned using the ultrachat_200k dataset to imbue it with conversational capabilities.

Key Characteristics & Purpose

• Contaminated Nature: The model is explicitly labeled as "TOTALLY CONTAMINATED," rendering it unreliable for general use.
• Reference Only: Its primary purpose is to serve as a reference point to highlight the vulnerabilities and potential unreliability of models, particularly in the context of leaderboard evaluations.
• Conversational Fine-tuning: Despite its compromised state, it underwent training on conversational data to demonstrate how such models might be constructed.

Important Considerations

• Do Not Use for Production: Due to its contaminated status, this model is not suitable for any practical application or deployment.
• Leaderboard Vulnerability: The creators emphasize that this model aims to illustrate how many models, even those on leaderboards, can be unreliable, urging users and sponsors to exercise caution.

This model is a unique case, designed not for performance but for educational and cautionary purposes regarding model integrity and evaluation.