repligate: breaks may help for 'information theoretical' reasons, not just neurobiology
repligate · x · 2026-10-27
Replying to @EpistemicHope's question about why humans benefit from breaks, repligate argues the reasons are "not merely neurobiological but also 'information theoretical'." This kicked off the thread in which he later claims that a context filled with noise is genuinely "hard" for LLMs.
Related event: AI Researchers Debate Whether LLMs, Like Humans, Benefit From Rest(5 posts)→
More from Models
- DeepSeek site shows new v4-flash fingerprints as community cracks checkpoint names — teortaxesTex · 2026-10-27
- User claims Kimi K3.1 Pro High generated an LLM explainer video; model name unverified — iamfakhrealam · 2026-10-27
- Anyone can get an abliterated model with safety training stripped out for pennies — EXM7777 · 2026-10-27
- Ask an LLM for 10 bold 2030 AI predictions — many already happened — FlorianGallwitz · 2026-10-27
- Prefilling probe shows curl trigger: 98% completions write a URL once '-o /tmp/' is appended — rickasaurus · 2026-10-27
- Unsloth details running DeepSeek-V4.1-Flash locally: 1-bit GGUF fits 128GB RAM with 1M context — gnukeith · 2026-10-27