Aggregate percentages tell you something improved. They don't tell you what. These are complete, unedited outputs from both models on the same held-out problems: the arithmetic the base model drops, and the step the tuned model doesn't.
The first five were selected because the two models disagree. That's what makes them worth reading, and it also means they are not a random sample. The tuned model scores 70.0%, not 100%, and the last card is a problem where both models fail. It's here on purpose: a gallery of nothing but wins is a highlight reel, not evidence.
James decides to run 3 sprints 3 times a week. He runs 60 meters each sprint. How many total meters does he run a week?
Kylar went to the store to buy glasses for his new apartment. One glass costs $5, but every second glass costs only 60% of the price. Kylar wants to buy 16 glasses. How much does he need to pay for them?
Eliza's rate per hour for the first 40 hours she works each week is $10. She also receives an overtime pay of 1.2 times her regular hourly rate. If Eliza worked for 45 hours this week, how much are her earnings for this week?
Toula went to the bakery and bought various types of pastries. She bought 3 dozen donuts which cost $68 per dozen, 2 dozen mini cupcakes which cost $80 per dozen, and 6 dozen mini cheesecakes for $55 per dozen. How much was the total cost?
Melanie is a door-to-door saleswoman. She sold a third of her vacuum cleaners at the green house, 2 more to the red house, and half of what was left at the orange house. If Melanie has 5 vacuum cleaners left, how many did she start with?
Toulouse has twice as many sheep as Charleston. Charleston has 4 times as many sheep as Seattle. If Seattle has 20 sheep, how many do the three have together?
Generated by demo/gen_traces.py from docs/sample_traces.md and demo/public/examples.json, both committed. Nothing here is edited for readability: the LaTeX fragments and the repetition are what the models actually emitted.