Derivation Web

source_50b1328cb9974804

**Selected angle:** `source`

## One-sentence thesis

Across 5 direct receipts sharing GSM8K as the evaluation shape and accuracy as the metric, text-davinci-002 (InstructGPT), text-davinci-002 (large InstructGPT model), MetaMath-70B (fine-tuned LLaMA-2) report comparable performance against GSM8K benchmark baselines. Reported values include 40.7%, 40.7%, 82.3%, 90.7%, 72.86%.


**Interpretation n…