NITP: Next Implicit Token Prediction for LLM Pre-training

Documenti analoghi

• Towards More Diverse and Challenging Pre-training for Point Cloud Learning: Self-Supervised Cross Reconstruction with Decoupled Views
  di: Zhang, Xiangdong, et al.
  Pubblicazione: (2025)
• PCP-MAE: Learning to Predict Centers for Point Masked Autoencoders
  di: Zhang, Xiangdong, et al.
  Pubblicazione: (2024)
• Put the Space of LoRA Initialization to the Extreme to Preserve Pre-trained Knowledge
  di: Tang, Pengwei, et al.
  Pubblicazione: (2025)
• Reasoning Bias of Next Token Prediction Training
  di: Lin, Pengxiao, et al.
  Pubblicazione: (2025)
• Implicit Optimization Bias of Next-Token Prediction in Linear Models
  di: Thrampoulidis, Christos
  Pubblicazione: (2024)