Back to Research
Scientific
Metric
Token-Entropy Analysis
Analyzing the relationship between structural noise ($N$) and signal density ($S$) across various optimization states.
Analysis Scenarios
Calculated Noise
($N_{opt}$)
$$N_{opt} = (H + D) \times (1 - \alpha)$$
Stupidity Index
($G$)
$$G = \frac{N_{opt}}{S + N_{opt}}$$
Metric Definitions
Signal
Useful Content
Tokens directly influencing LLM
utility.
Noise
Entropy Load
Excess tokens (Structure/Meta/Ads).