Prove it
outside

Ask an open model which of two places is quieter. Lock its answer. Go there and measure.

Place A
Place B

1 Describe the places

Every model gets this exact text, so any difference comes from the model.

Place A Place B When you will measure

Fill in both places, then build the prompt.

Run the prompt on each open model, for example with run_models.py, then paste the answers in step 2.

2 Lock the predictions

Name of place A Name of place B

Run the prompt from step 1 on each open model (any chat or runner works), then enter what each one answered. Confidence is the percent chance the model gave that its pick is right.

Paste JSON from run_models.py instead

Predictions (JSON array)

Not locked yet. After locking, the predictions cannot change.

3 Measure

Conditions

Run each place at least twice. Same phone, same posture, screen on, stand still.

No microphone? Analyze a recording instead

Use a recording of each place, at least 30 seconds long. Each full segment (up to two per file) becomes a run. Shorter segments are less reliable; say so in your write-up. The source is saved with the results and shown in the score. Levels are only comparable when both recordings were made with similar equipment and settings.

Audio file Where is this recording from? (title, author, license, link) Segment length

Waiting for a measurement.

4 Score

Not scored yet.