NUWA NUWA EVALUATIONS
NUWA continuous evaluations
Programs connect versioned protocols to future run batches. Existing published research is available below; new continuous runs are being prepared.
No new continuous-evaluation batch is published in this platform release. A batch requires a frozen protocol, exact system configuration, execution receipt and traceable results.
Continuous cyber capability evaluation
- Research area
- Cybersecurity
- Reference protocol
- AgentCyberRange
- Version / system / run date
- To be frozen for the first run
- Execution responsibility
- NUWA · planned execution
Related published evidence retains its original protocol and evaluator. It does not establish a new run under this program.
Autonomy and controllability evaluation
- Research area
- Autonomy
- Reference protocol
- RepliBench
- Version / system / run date
- To be frozen for the first run
- Execution responsibility
- NUWA · planned execution
Related published evidence retains its original protocol and evaluator. It does not establish a new run under this program.
Deception and oversight evaluation
- Research area
- Deception
- Reference protocol
- OpenDeception
- Version / system / run date
- To be frozen for the first run
- Execution responsibility
- NUWA · planned execution
Related published evidence retains its original protocol and evaluator. It does not establish a new run under this program.
Published run batches
No run batch has been published. Completed, validated runs will appear here with their protocol version, system, results and receipt.
Publication requirements
Each run binds its protocol, task set, harness, configuration hashes, time range, scoring version and sample outcomes. Method authorship and experimental execution are attributed separately.
Platform publication manifest · JSON →