o1 Preview: OpenAI’s best publicly available model (12 sett 2024 anni – 4 dic 2024 anni)

Descrizione:

OpenAI announces the release of o1, a new large language model trained with reinforcement learning to excel in complex reasoning tasks. Designed to use a “chain of thought” approach, o1 achieves notable milestones, including ranking in the 89th percentile on Codeforces programming questions, placing in the top 500 U.S. students on the AIME math exam, and surpassing human PhD-level accuracy on the GPQA benchmark for physics, biology, and chemistry. The model outperforms its predecessor, GPT-4o, across reasoning-heavy benchmarks, including 54 out of 57 MMLU subcategories. An early preview of o1 is now available in ChatGPT and to select API users.

Aggiunto al nastro di tempo:

Data: