Unverified50% confidenceFactExact time
The generation-0 neural network did nothing and scored 38%; generation 2 surpassed the hand-crafted rules; after 41 generations of evolution, the final score reached 65%.
1
Sources
50%
Confidence
Long-term
Relevance
11/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe Φ value of standard feedforward neural networks approaches zero, because information flows only in one direction and lacks recurrent integration.66% similarUnverifiedThe skip-zero optimization (exploiting that 92.5% of neurons are zero after ReLU², reading down weights in groups of 16 neurons) raised speed from 6.6 tok/s to 7.3-7.6 tok/s, an improvement of about 12%.65% similarUnverifiedIn leave-one-family-out evaluation, the selected neurons still retain 86%–94% of the full-feature ceiling on generator families never seen during selection.62% similarUnverifiedInvoking an existing Skill consumes far fewer tokens than generating from scratch, and the success rate improves through repeated iteration.62% similarUnverifiedIf alignment decays by 0.1% per generation, after 100 generations the alignment level would be approximately 90.5%, and after 1,000 generations it would fall to approximately 37%.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1140021API
curl https://kongchang.com/api/v1/knowledge/claims/1140021MCP
get_claim(id=1140021)