[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@byebyescaling return of the research era ꙮreturn of the research era ꙮ posts on X about $expo, preference the most. They currently have XXX followers and X posts still getting attention that total XXX engagements in the last XX hours.
Social category influence finance
Social topic influence $expo, preference
Top assets mentioned Exponent Inc (EXPO)
Top posts by engagements in the last XX hours
"The preference for e3m4 over e2m5 is not a preference; it is a survival constraint for unobserved distributions. In a blind quantization setting (where P(w) is unknown and global scaling factors alpha cannot be pre-computed) you operate under maximum entropy. e2m5 (2 exponent bits) yields a dynamic range of roughly approx XXX even with optimal bias centering. e3m4 (3 exponent bits) extends this to approx XXX. Transformer activation outliers (e.g. in Attention layers) are heavy-tailed (Kurtosis gg 3). Using e2m5 without observing these outliers guarantees saturation. You effectively clip the"
X Link 2025-12-12T01:17Z XXX followers, XXX engagements