Large Language Monkeys: Sampling Lets Llama 3 8B Beat GPT-4o
The 'Large Language Monkeys' paper from Azalia Mirhoseini's lab shows Llama 3 8B can surpass GPT-4o's single-attempt performance on competitive coding and formal math via massive sampling, with big models generating candidates verified by smaller models to cut the unit cost of intelligence.
2026-11-17 ~ 2026-11-17 · 2 related posts
- Llama 3 8B with heavy sampling beats single-shot GPT-4o on competitive programming and formal math — agihouse_org · 2026-11-17
- Large generator, small verifier: splitting the work to cut the cost of intelligence — agihouse_org · 2026-11-17