Increase DeepSeek V4 Flash weight_scale_factor to 0.246
Gonka proposal #98: Increase the weight_scale_factor for deepseek-ai/DeepSeek-V4-Flash-0731 from 0.214 to 0.246. All other model and chain parameters remain...
Yes
100%
No
0%
Abstain
0%
Veto
0%
Original Proposal Text
Increase the weight_scale_factor for deepseek-ai/DeepSeek-V4-Flash-0731 from 0.214 to 0.246. All other model and chain parameters remain unchanged. Resubmission of proposal 97, which expired without reaching quorum. No votes were cast against it. The content is unchanged. The original factor came from inflated nonce/min numbers. The measurement fix and reproduction steps are in gonka-ai/gonka#1640 (open, will be proposed with the next mainnet update). Recomputed on historical B300 nonce data and fresh PoC experiments, the factor is 0.246. Parameters: epoch_params: [object Object] validation_params: [object Object] poc_params: [object Object] tokenomics_params: [object Object] collateral_params: [object Object] bitcoin_reward_params: [object Object] dynamic_pricing_params: [object Object] bandwidth_limits_params: [object Object] confirmation_poc_params: [object Object] genesis_guardian_params: [object Object] developer_access_params: [object Object] participant_access_params: [object Object] transfer_agent_access_params: [object Object] devshard_escrow_params: [object Object] fee_params: [object Object] delegation_params: [object Object] maintenance_params: [object Object]