Trust
Trust is a Bittensor consensus metric for the miner side of evaluation. It describes how much miner support remains after validator weights are clipped by consensus (Glossary: Trust, Yuma Consensus).
The term belongs to Yuma Consensus vocabulary because it is read after validator weight signals have been filtered and aggregated. It is not a standalone miner reputation label.
In plain language, trust answers: how much validator support did this miner retain after consensus clipping removed outlier weight signals? A miner with high trust received broad support that survived the consensus filtering step, not just a few validators rating it highly (Glossary: Trust).
For a reader, trust differs from raw validator weight totals. Raw weight totals count all support before filtering; trust reflects the miner-side support that consensus considers reliable after comparing validator signals across the set (Yuma Consensus).
Trust therefore sits after validator weights and consensus score. Validator weights provide input, consensus score supports clipping, and trust describes the miner-side support left after that filtering step (Glossary: Validator Weights, Glossary: Consensus Score).
Consensus Clipping
Validators set weight signals after evaluating miner work. Consensus filtering can reduce outlier support before the final miner-side result is read (Consensus Score, Glossary: Trust).
Trust describes the miner-side effect of that filtering. It is therefore about consensus-adjusted support, not a separate label outside the consensus process.
Trust describes what remains of miner-side support after consensus filtering. Validator weights name the original input signals, and emissions describe later reward flow (Consensus Score, Glossary: Trust, Emission).
Rank and Pre-Rank
Rank and trust are connected but not identical. Rank names the consensus-adjusted miner result, while trust describes how much support is preserved between pre-rank and final rank (Glossary: Rank, Glossary: Trust).
Rank says where the miner lands after aggregation. Trust says how much clipping changed the pre-rank-to-rank relationship.
That relationship keeps the result and the comparison separate. Rank is the adjusted miner-side outcome, while trust describes the preservation of support between the pre-clipping signal and final rank (Glossary: Rank, Yuma Consensus).
That makes trust a comparison term inside the Yuma process. It is meaningful because pre-rank and rank can differ after clipping.
Validator Trust
Trust is read from the evaluated miner’s side of the process. Validator trust is read from the validator side, describing validator influence after clipping (Glossary: Validator Trust, Glossary: Trust).
Keeping the two terms separate avoids treating trust as a universal quality score for every role in a subnet.
The distinction is role direction. Trust concerns support for an evaluated miner. Validator trust concerns validator-side influence reflected by clipped weights (Glossary: Validator Trust, Glossary: Validator Weights).
Weights and Scores
Validator weights are the input signals, and consensus score is the benchmark used to clip outlier weights. Trust is read after those filtering steps affect miner-side support (Glossary: Validator Weights, Glossary: Consensus Score).
This makes trust downstream of weight input language. A validator weight can contribute to the process, but trust describes the miner-side relationship after consensus filtering.
Weight matrix terminology is also earlier than trust. The matrix collects validator weight vectors, while trust is interpreted after the Yuma process has filtered and aggregated those signals (Glossary: Weight Matrix, Glossary: Trust).
This order keeps input and result terms separate. Weights and matrices describe submitted evaluation material; trust describes a miner-side result after filtering.
Same-Subnet Comparison
Trust comparisons are most meaningful when the compared values come from the same subnet consensus setting. Yuma Consensus operates over validator weights inside a subnet, so miner-side trust values are clearest when read against the process that produced them (Yuma Consensus, Consensus Score).
Across different subnets, the task, validator set, incentive mechanism, and weight pattern can differ. Trust keeps the same definition, but the surrounding consensus setting can change what a particular value is telling the reader.
A trust value is most informative when the subnet task, validator set, and consensus process remain part of the comparison (Yuma Consensus, Understanding Incentive Mechanisms).
Incentive Role
Trust belongs near emissions vocabulary because miner-side support, rank, and clipping all feed into the broader Yuma Consensus path for incentives (Yuma Consensus, Emission).
Trust explains a consensus relationship between pre-rank and rank, while emissions describe the broader reward-flow setting.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For Trust, that sequence changes how an example should be read, because the surrounding network state differs at each stage.
In localnet, trust can be exercised in an isolated development environment, where the surrounding chain state reflects local configuration rather than production history.
On testnet, trust can be observed in a shared, non-production network whose state is kept separate from mainnet.
On mainnet, trust applies on the live, production Bittensor network, where the surrounding state is real and persistent.
The Bittensor Networks reference separates mainnet, testnet, and localnet. A trust example from one environment should not be read as representing another environment.
Relationship to Yuma Consensus
Trust and Yuma Consensus describe related parts of Bittensor’s incentive system. Yuma Consensus is the on-chain process that aggregates validator weight signals within a subnet into miner incentives and validator dividends, applying consensus clipping, bonding, and emission calculation (Yuma Consensus).
For readers, trust names a specific part of that incentive picture, while Yuma Consensus names the consensus process that turns validator weights into the resulting incentives and dividends.
Reader Boundary
Trust is a miner-side consensus metric after clipping. It is not rank itself, validator trust, or a reward guarantee (Glossary: Trust, Yuma Consensus).
Specific trust values belong to the subnet, network, and consensus run that produced them.
Consensus Score Sets the Clipping Benchmark
The Glossary: Consensus Score names the stake-weighted benchmark Yuma Consensus uses when validator weights are combined into final rank results. Yuma Consensus describes weights that exceed that benchmark as candidates for clipping before aggregation continues.
Trust is read after that filtering step on the miner side. Consensus score names the agreement benchmark; trust describes how much miner-side support remains once outlier validator signals have been reduced (Glossary: Trust).
Trust Compares Pre-Rank Support to Final Rank
The Glossary: Trust describes trust as a measure of how much miner-side support remains after consensus clipping, comparing support before and after the final rank result is produced. A miner can receive strong raw validator weights yet see trust fall if many of those signals were treated as outliers relative to the consensus benchmark.
That comparison keeps trust distinct from rank itself. Rank names where the miner lands after aggregation; trust names how much support survived the clipping step that shaped that result (Glossary: Rank).
Validator Trust Reads the Validator Side Instead
Validator trust measures how consistently a validator’s weights align with consensus over time on the validator side of the process. Trust instead stays on the evaluated miner’s side and describes retained support after clipping.
The two terms share one Yuma pass but point in different directions. Validator trust concerns validator-side influence after filtering; trust concerns miner-side support preservation after the same filtering step (Yuma Consensus).
Network Examples
Trust examples from one environment are not evidence for another (Bittensor Networks).
Relationship to Validator Trust
Trust and validator trust are related but different Yuma Consensus metrics. Trust names miner-side support after validator weights are clipped by consensus, while validator trust names validator-side influence after that same clipping step (Glossary: Trust, Glossary: Validator Trust).
For readers, trust belongs to miner evaluation and validator trust belongs to validator influence. Both are read after outlier validator weights are filtered, but one summarizes retained miner support and the other summarizes how much validator signal remains influential. They mirror opposite sides of the same consensus process rather than naming the same role.
Both terms should stay inside Yuma Consensus vocabulary instead of being read as general reputation labels. A strong trust reading for a miner does not automatically imply a matching validator trust reading, because the two metrics answer different questions in the same pipeline (Yuma Consensus).
References: Glossary: Trust, Glossary: Validator Trust, Yuma Consensus