On Next-Token Prediction in LLMs: How End Goals Determine the Consistency of Decoding Algorithms

Documenti analoghi

• Characterizing the Multiclass Learnability of Forgiving 0-1 Loss Functions
  di: Trauger, Jacob, et al.
  Pubblicazione: (2025)
• Predictive Concept Decoders: Training Scalable End-to-End Interpretability Assistants
  di: Huang, Vincent, et al.
  Pubblicazione: (2025)
• Efficient Training of Language Models with Compact and Consistent Next Token Distributions
  di: Sathe, Ashutosh, et al.
  Pubblicazione: (2024)
• Reasoning Bias of Next Token Prediction Training
  di: Lin, Pengxiao, et al.
  Pubblicazione: (2025)
• ENTP: Encoder-only Next Token Prediction
  di: Ewer, Ethan, et al.
  Pubblicazione: (2024)