Skip to main content

Fairness Mechanisms

Resultity includes multiple layers of fairness enforcement to maintain trust in a decentralized network where nodes are operated independently by users. These mechanisms balance openness with accountability, ensuring that job execution is honest, consistent, and useful.

Blind Jobs​

The network regularly injects blind test jobs into the queue. These jobs:

  • Are indistinguishable from regular inference tasks;
  • Have known expected outputs or statistical fingerprints;
  • Are assigned randomly across the node pool;
  • Are not linked to any real client or billing.

Blind jobs are used to:

  • Assess output quality and correctness;
  • Detect manipulated or low-effort execution;
  • Establish per-node quality scores over time.

Failure to perform accurately on blind jobs results in penalties or temporary exclusion from job rotation.

Response Validation​

Each inference result may be subject to:

  • Automated validation: output structure, token count, latency patterns, GPU consistency;
  • Manual review: randomly sampled logs, operator audits (especially during testnet);
  • Delayed comparison: cross-checking multiple responses from replicated jobs.

The network maintains response logs and aggregates historical data to track anomalies or regressions.

Feedback and Reporting​

Clients (developers or integrated systems) may optionally:

  • Provide structured feedback (e.g., thumbs up/down, categorization);
  • Flag problematic responses for review;
  • Participate in reward adjustments (future DAO voting mechanisms).

Feedback is not required but contributes to long-term trust scoring.

Penalties and Exclusion​

Nodes may be temporarily or permanently excluded from job rotation for:

  • Repeated failed jobs;
  • Invalid or missing signatures;
  • Unresponsive behavior;
  • Systematic quality degradation.

Penalized nodes receive notifications (via WebSocket or dashboard) and may request reinstatement after cooldown or manual verification.

Node Scoring (Planned)​

Future iterations may include:

  • Public or semi-public scoring dashboards;
  • Weight-based rotation (better scores = more job opportunities);
  • Decentralized voting on node behavior and reward eligibility.

These tools will be governed either by protocol rules or community-voted governance frameworks.

Resultity’s fairness logic is designed to preserve decentralization while discouraging manipulation, rewarding reliability, and enabling scalable trust without direct supervision.