monad.ai / WEIGHT_MIN
Variable: WEIGHT_MIN
constWEIGHT_MIN:0.01=0.01
Defined in: kernel/adaptiveWeights.ts:33
Hard floor on any learned weight.
No scorer can drop below 1% influence regardless of how many negative rewards it accumulates. This ensures every signal remains recoverable: if a scorer later becomes genuinely useful, it can climb back up.