Forcing LLMs to be evil during training can make them nicer in the long run
"模型训练" 相关文章
Evaluating the Dynamics of Membership Privacy in Deep Learning
G-Core: A Simple, Scalable and Balanced RLHF Trainer
Adaptive Multimodal Protein Plug-and-Play with Diffusion-Based Priors
Improving Task Diversity in Label Efficient Supervised Finetuning of LLMs
Predicting Brain Responses To Natural Movies With Multimodal LLMs
Studying Cross-cluster Modularity in Neural Networks
ASR-Guided Speaker-Role Diarization and Diarization-Guided ASR Decoding
A new study just upended AI safety
[程序员] 为什么没有平台，可以连接世界上任何电脑的显卡，整合算力进行训练
[程序员] 为什么没有平台，可以连接世界上任何电脑的显卡，整合算力进行训练
A Simple "Try Again" Can Elicit Multi-Turn LLM Reasoning
The Impact of Language Mixing on Bilingual LLM Reasoning
QK-Clip巧解MaxLogit爆炸难题：让Muon在Scaleup之路上更进一步
大模型落地基础技术体系LLM<RAG<AI Agent<Training
There's no way to stop models knowing they've been rolled back
Mixture of Raytraced Experts
EASTER: Embedding Aggregation-based Heterogeneous Models Training in Vertical Federated Learning
VIVID-10M: A Dataset and Baseline for Versatile and Interactive Video Local Editing
Dataset Distillation-based Hybrid Federated Learning on Non-IID Data
