| 🚀 Qwen2.5-3B Fine-Tuned on BBH (Formal Fallacies) - Model Card | |
| 📌 Model Overview | |
| Model Name: Qwen2.5-3B Fine-Tuned on BBH (Formal Fallacies) | |
| Base Model: Qwen2.5-3B-Instruct | |
| Fine-Tuned Dataset: BBH (BigBench Hard) - Formal Fallacies | |
| Task: Logical Reasoning & Deductive Validity Classification | |
| Fine-Tuning Objective: Improve the model’s ability to classify logical arguments as valid or invalid based on deductive reasoning principles. | |
| 📌 Dataset Information | |
| This model was fine-tuned on the Formal Fallacies subset of the BigBench Hard (BBH) dataset. | |
| Dataset characteristics: | |
| Task Type: Deductive reasoning & formal logic classification | |
| Input Format: Logical argument statements presented in natural language | |
| Target Labels: "valid" or "invalid" | |
| Example: | |
| Input: | |
| "Here comes a perfectly valid argument: First, being a cousin of Chris is sufficient for not being a son of Kermit. We may conclude that whoever is not a son of Kermit is a cousin of Chris." | |
| Target: "invalid" | |
| This dataset evaluates a model’s ability to identify logically valid vs. invalid arguments, which is crucial for AI-assisted legal analysis, debate systems, and automated theorem proving. |