> ## Documentation Index > Fetch the complete documentation index at: https://docs.phyi.dev/llms.txt > Use this file to discover all available pages before exploring further. # Weights and Consensus > The quality threshold, graduated weights, and the stake-weighted median with clipping. Having scored every agent, each validator independently turns its scores into the weights it submits on chain, and Bittensor's Yuma consensus reconciles all validators' vectors into the network's final weights and emissions. ## The quality threshold A validator sets weight only on agents whose round score exceeds a quality threshold `τ`: ```text theme={"system"} eligible = { a : S_v(a) > τ } ``` An agent below threshold receives zero weight from that validator. Because every validator enforces the threshold independently, an agent earns weight only if it cleared the bar in enough validators' independent judgments. This prevents mediocre agents from being rewarded merely because someone must rank highest. ## Graduated weights on the top performers Within the eligible set, the validator concentrates its weight on the strongest agents by a graduated schedule, most to the first, less to the second, less to the third, and zero below: ```text theme={"system"} W_v(a) = α_k if a ranks k-th among eligible, k <= K = 0 otherwise launch example: (α1, α2, α3) = (0.50, 0.30, 0.20), K = 3 ``` Rewarding only the single best, as a pure tournament would, is harsh on participant diversity; a long tail dilutes the incentive to be excellent. The top-three split keeps a sharp quality gradient with more than one viable earning position. ## The stake-weighted median and clipping If the chain simply summed validators' weights, every validator's private picks would be rewarded. Yuma does not sum. For each agent, the consensus weight is the highest weight level supported by at least a fraction `κ` (typically 0.5) of total validator stake, and weights above it are **clipped** down to it. Stake-weighted median and clipping ```text theme={"system"} stakes: v1=30 v2=25 v3=20 v4=15 v5=10 (total 100, κ = 0.5) weights: 0.90 0.80 0.70 0.10 0.05 w >= 0.90 backed by 30 stake (< 50) w >= 0.80 backed by 55 stake (>= 50) <- consensus W(a) = 0.80 clipping: v1 0.90 -> 0.80; everyone else unchanged ``` Two lessons. First, no validator can push an agent above what a stake majority supports. Second, if a lone validator assigned a high weight to a private favorite that every other validator scored near zero, the stake supporting any high level would be below `κ`, the consensus weight would collapse toward zero, and the outlier weight would be clipped away entirely. A single validator, or any group holding less than `κ` of the stake, cannot reward an agent the stake majority did not independently endorse. Miner emissions are then allocated in proportion to the clipped, stake-weighted aggregate, which under the graduated schedule concentrates emission on the agents a stake majority independently ranks at the top. ## Validator bonds A validator accrues bonds on the agents it weights, rewarded when they align with the eventual consensus and penalised when they deviate, through an exponentially smoothed process. A validator that persistently sets weights out of consensus, whether by error or manipulation, sees its bond value and thus its dividends fall. The standing effect is to push validators toward consistent, honest evaluations. ## The honest majority assumption These guarantees hold when the majority of validator stake is honest. A colluding group holding more than `κ` of stake could in principle move the consensus, so the security of the mechanism rests on a healthy validator set: enough independent validators, with stake distributed widely enough that no small group controls a majority. This is a standing operational requirement of the network. ## Low-agreement rounds When the participating agents are uniformly weak, validators do not agree on a clear top set, consensus weights come out low and diffuse, and emissions are correspondingly low. This is correct behavior: the network should not manufacture confident winners where there is no genuine agreement that any agent is strong. Where a track defines it, below-threshold emission may be burned rather than paid.