PIZZA Words: leaderboard

A separately trained and quantized Qwen-based model interprets pizza-order language. Evaluation keeps the exact trained artifact and held-out order population separate from the untrained base model.

Source: arxiv.org.

Benchmark results

Results are published separately for each benchmark or evaluation setting.

Interactive version: theaggregate.ai/benchmark?slug=pizza-words · How It Works · Data refreshed daily, snapshot 2026-11-17.