News roundup 44

Created on December 26|Last edited on December 27
Comment
xAI Raises $6 Billion, Eyes $50 Billion Valuation
Elon Musk's AI startup secures massive funding, aiming to compete with OpenAI and Anthropic through its Grok model.
﻿https://wandb.ai/byyoung3/ml-news/reports/xAI-Quietly-Raises-Another-6-Billion--VmlldzoxMDczMzk1NA﻿
DeepSeek-V3: 671 Billion Parameters Trained on a $6 Million Budget
New MoE model outperforms competitors on multiple benchmarks with unprecedented efficiency, challenging the need for massive GPU clusters.
﻿https://api.wandb.ai/links/byyoung3/tbkwjmdl﻿
OREO Revolutionizes Multi-Step Reasoning in LLMs
Offline RL method boosts reasoning performance, achieving state-of-the-art results on GSM8K and ALFWorld while cutting dependency on costly pairwise data.
﻿https://arxiv.org/abs/2412.16145﻿
Tencent’s DRT-o1 Sets New Benchmark for Machine Translation with Long Chain-of-Thought  
Using a multi-agent framework, Tencent’s DRT-o1 integrates advanced reasoning processes into neural machine translation, achieving an unprecedented 8.26 BLEU improvement on literature translation tasks.
﻿https://arxiv.org/pdf/2412.17498v1﻿
﻿
Add a comment