Antix5/vi-gym-causal-ascii

Vi-Gym Causal ASCII Trajectories

This dataset contains autoregressive trajectories of a Large Language Model (LLM) agent learning spatial reasoning and geometric drawing within a simulated Vi (Vim) editor environment.

Warning

This dataset is a direct derivation of the source material, it might therefore also contain content not suitable for all audiences. All authors of the original artwork have full ownership.

Dataset Structure

Each record is a discrete step in the environment, capturing the exact state of the editor before a command is issued. The format is designed for Causal Next-Token Prediction training.

Format Specification

<BOS>
<notepad>
[CURRENT ASCII CONTENT]
</notepad>
<mode>[Normal|Insert]</mode>
<prompt>[Grammatically Correct Instruction]</prompt>
<command>
[OPTIMIZED VI KEYSTROKES]

Technical Provenance

1. 1.Environment Engine: States rendered via the Rust-based Vi-Gym engine.
2. 2.Keystroke Optimization: Generated using an AST-based compiler prioritizing efficiency via Run-Length Encoding (RLE) and geometric entropy sorting.
3. 3.Linguistic Robustness: Prompts utilize grammatically correct indefinite articles (a/an) and randomized natural language templates with human-like noise.

Source Credits

• —Geometric Data: ASCII art shapes sourced from the Curated ASCII Art Database (originally from asciiart.eu).
• —Logic Backend: Editor state and command interpretation powered by the ViLM core engine (not published yet).

Training Recommendations

Calculate loss only on tokens following the <command>\n tag to focus the model on the mapping between visual state and command execution.