Baidu released ERNIE 5.5 Flash Base, a sparse Mixture-of-Experts language model with 280B total parameters and 14B activated per token, pre-trained from scratch on 33.5T tokens with a 128K training context and 1M-token support for fine-tuning and inference. The release is the end-of-pre-training checkpoint before any mid- or post-training, positioned as a common starting point for domain adaptation, instruction tuning and reinforcement-learning research.
Base Model
基座模型
Pre-trained large model with general capabilities, not fine-tuned for specific tasks
Curated Intelligence
A new Nvidia paper reveals that scoring base models at every step of an agent run is ineffective, whereas evaluating them only on the specific edit that first makes tests pass accurately predicts post-training performance. By replaying successful trajectories to identify this decisive step, the method achieves a 0.964 rank correlation with SWE-bench Verified scores, offering a low-cost way to select base models for agentic fine-tuning.
The Hugging Face Open R1 project released its fourth update, officially integrating DeepSeek-V3-0324 as the base model. While maintaining the open-source reasoning training pipeline, this version addresses community concerns with a detailed safety assessment, confirming that downloading and running the models is safe but emphasizing the need for human-in-the-loop verification in agentic applications to mitigate generation risks.