cris177/Qwen2-Simple-Arguments
cris177/Qwen2-Simple-Arguments is a 0.5 billion parameter Qwen2-based language model developed by Cristian Desivo, fine-tuned for parsing simple English arguments. It specializes in identifying premises, conclusions, propositions, argument types, negations, and validity from two-premise, two-proposition arguments. The model is optimized for logical analysis tasks, providing structured JSON output for argument components.
Loading preview...
Overview
cris177/Qwen2-Simple-Arguments is a 0.5 billion parameter Qwen2-based language model developed by Cristian Desivo. It is specifically fine-tuned to analyze and parse simple English arguments, which are defined as arguments consisting of two premises and a conclusion, involving two distinct propositions. The model's primary function is to deconstruct these arguments into their logical components and assess their validity.
Key Capabilities
- Argument Parsing: Identifies and extracts premises, conclusions, and individual propositions from simple English arguments.
- Logical Analysis: Determines the type of argument (e.g., Modus Ponens, Modus Tollens, Affirming the Consequent, Disjunctive Syllogism, Denying the Antecedent, Invalid Conditional Syllogism).
- Proposition Negation: Generates the negation of each identified proposition.
- Validity Assessment: Evaluates and outputs the logical validity of the argument as a boolean.
- Structured Output: Provides analysis in a structured JSON format, making it suitable for programmatic use.
Training Details
The model was trained on a synthetic dataset of 50,000 arguments, covering various argument types. This specialized training ensures high accuracy for its intended task, achieving 100% accuracy on its synthetic train and test sets. Training was performed using unsloth for efficiency, requiring less than 2.5 GB of VRAM and completing in 2.5 hours for one epoch.
Good For
- Automated logical argument analysis.
- Educational tools for logic and critical thinking.
- Applications requiring structured extraction of argument components from text.