Thermodynamics of words: word frequencies in literature match Rayleigh-Jeans thermalization
burny_tech · x · 2026-11-13
A playful physics preprint, "Thermodynamics of words," proposes an "word thermalization hypothesis": words in literary texts have effective interactions, and their frequency distributions take the same form as Rayleigh-Jeans thermalization and condensation — treating word statistics like a thermodynamic system where high-frequency words behave like a condensate.
More from Research
- Audio Team Built LiquidCodec, a 74M-Parameter Streaming Audio Codec — helloiamleonie · 2026-11-13
- Memgraph Fine-Tuned a 1B Model for Text2Cypher with $26 of Synthetic Data: 94% Synthetic, 41% Public Benchmark — JeremyCMorgan · 2026-11-13
- 95-99% of Gradient Norm Lost at LM Head; Raising Head Rank 32→4096 Speeds Convergence 16x — yoavartzi · 2026-11-13
- BioSecBench-Structure: 58 evaluations benchmarking agents' molecular reasoning for biosecurity — kenbwork · 2026-11-13
- Daniel Kokotajlo: pretraining loss is a checkable task proxying nearly all downstream skills — nitarshan · 2026-11-13
- MIT releases AI Risk Mitigation Taxonomy v3: 2,206 strategies from 47 frameworks — Exp_Mark · 2026-11-13