pub struct BenchTrustSummary {
pub suite_id: String,
pub bench_trust_score: f64,
pub total_tasks: u32,
pub flagged_task_count: u32,
pub audited_at: DateTime<Utc>,
pub anchor: String,
}Expand description
Summary of a bench audit produced by the Python eval plane.
Field semantics mirror fd_evals.bench_audit.BenchAuditReport — only the
subset the Rust policy plane needs to decide is included here. The full
per-task flag list lives in the eval-plane report; this struct intentionally
stays narrow so it can be denormalised onto a runs / eval_runs column
without bloating the row.
Fields§
§suite_id: StringIdentifier of the eval suite that was audited.
bench_trust_score: f64Aggregate hygiene score in [0.0, 1.0]. Higher = more trustworthy.
total_tasks: u32Total tasks the auditor scanned in the suite.
flagged_task_count: u32Number of distinct tasks with at least one hygiene flag.
audited_at: DateTime<Utc>When the audit was produced.
anchor: StringarXiv anchor of the audit methodology (defaults to
BENCH_AUDIT_ANCHOR).
Implementations§
Source§impl BenchTrustSummary
impl BenchTrustSummary
Sourcepub fn flagged_task_ratio(&self) -> f64
pub fn flagged_task_ratio(&self) -> f64
Share of suite tasks with at least one hygiene flag. The policy plane
uses this as the noise margin around a benchmark delta — if the cited
improvement is within flagged_task_ratio, it’s not distinguishable
from grading noise and the gate must deny.
Trait Implementations§
Source§impl Clone for BenchTrustSummary
impl Clone for BenchTrustSummary
Source§fn clone(&self) -> BenchTrustSummary
fn clone(&self) -> BenchTrustSummary
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
source. Read more