N-Gram House

Tag: LLMs

Positional Encoding in Transformers: Sinusoidal vs Learned for LLMs

Positional Encoding in Transformers: Sinusoidal vs Learned for LLMs

Sinusoidal and learned positional encodings were early solutions for transformers, but modern LLMs now use RoPE and ALiBi for better long-context performance. Learn why and how these techniques evolved.

Categories

  • History (50)
  • Machine Learning (30)

Recent Posts

Scheduling Strategies to Maximize LLM Utilization During Scaling Jan, 6 2026
Scheduling Strategies to Maximize LLM Utilization During Scaling
Architectural Innovations Powering Modern Generative AI Systems Nov, 7 2025
Architectural Innovations Powering Modern Generative AI Systems
Guardrail-Aware Fine-Tuning to Reduce Hallucination in Large Language Models Feb, 1 2026
Guardrail-Aware Fine-Tuning to Reduce Hallucination in Large Language Models
Automated Architecture Lints: Enforcing Boundaries in Vibe-Coded Apps Jan, 26 2026
Automated Architecture Lints: Enforcing Boundaries in Vibe-Coded Apps
How Design Teams Use Generative AI for Wireframes, Creative Variations, and Asset Generation Jan, 21 2026
How Design Teams Use Generative AI for Wireframes, Creative Variations, and Asset Generation

Menu

  • About
  • Terms of Service
  • Privacy Policy
  • CCPA
  • Contact

© 2026. All rights reserved.