VIDRAFT has released its Darwin-180B-RSI model, a 180-billion-parameter reasoning model that has achieved top rankings on five Hugging Face leaderboards. The model notably secured perfect scores on the AIME and HMMT benchmarks, marking a significant milestone. It employs proprietary techniques including selective model merging, recursive self-improvement, and zero-token confidence to enhance its capabilities in complex STEM-related reasoning tasks. AI
IMPACT Sets new SOTA on math and science reasoning benchmarks, potentially influencing future development in specialized STEM AI.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
