iLB is a production-grade, unbounded procedural benchmark for AI logic evaluation.

🇨🇳 中文文档 (Chinese Documentation)

• Unbounded Difficulty: No max score.
• Language Drift Metrics: Does logic hold across languages?
• Isomorphic Consistency: Truth Index (Ti).

pip install -e .

ilb

from ilb import api
# Generate a task
task = api.generate_task(engine="RecursiveFunction", difficulty=5)
print(task.prompt)

• Core: Schema definitions.
• Engines: 30+ logic engines across Math, Code, Causal, and Meta domains.
• Ice: Rendering layer.
• Runner: Execution logic.
• UI: Textual-based TUI.