You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
REP-005 — Drive and validate CAGE-2 replication across the CybORG sim and the ACES reference backend #639
Use one CAGE-2 SDL scenario (REP-003) with both the CybORG simulator adapter (REP-004) and the ACES reference backend. Capture reproducible, paper-ready evidence that both realize the authored scenario consistently under the equivalence rules in REP-001.
Done when
The same SDL revision drives both supported paths.
The evidence records each backend, configuration, result, limitation, and the REP-001 comparison.
Another researcher can repeat the reported result.
CAGE-2 emulation-backend realization is outside this issue. Requirement: REP-005. Design: #635. Milestone: Paper Slice: CAGE-2 Replication (ACES-Driven Sim) (#58).
The currently retained result does not satisfy REP-005:
-3112.2 is the published CAGE-2 README leaderboard total for CCS Sleeper.
-3112.6716 is the observed primary total from the 9,000 valid OpenRAE-mediated runs in the same-CybORG behavioral reconstruction (-3112.6715999999997 in the retained binary64 aggregate).
The frozen comparison classifies that result as bounded.
The supported claim is bounded behavioral-baseline outcome reproduction. It is not evidence of two independent conformant backends, a RAES reference-backend realization, direct-native-versus-OpenRAE equivalence, submitted-agent artifact identity, or state/observation equivalence.
changed the title [-]REP-005 — Drive and validate CAGE-2 replication across the CybORG sim and the ACES reference backend[/-][+]Deferred: CAGE-2 cross-backend replication (REP-005)[/+]on Sep 20, 2026
changed the title [-]Deferred: CAGE-2 cross-backend replication (REP-005)[/-][+]REP-005 — Drive and validate CAGE-2 replication across the CybORG sim and the ACES reference backend[/+]on Sep 20, 2026
Activity
doublewhy commentedon Aug 13, 2026
Evidence and claim correction (2026-08-13)
The currently retained result does not satisfy REP-005:
-3112.2is the published CAGE-2 README leaderboard total for CCS Sleeper.-3112.6716is the observed primary total from the 9,000 valid OpenRAE-mediated runs in the same-CybORG behavioral reconstruction (-3112.6715999999997in the retained binary64 aggregate).bounded.The supported claim is bounded behavioral-baseline outcome reproduction. It is not evidence of two independent conformant backends, a RAES reference-backend realization, direct-native-versus-OpenRAE equivalence, submitted-agent artifact identity, or state/observation equivalence.
Evidence: OpenRAE/adapters#91 and the frozen aggregate at the retained research head in OpenRAE/research#75. The stronger paired direct-native-versus-mediated experiment is separately tracked by OpenRAE/research#76.
Keep #639 open until its actual two-backend acceptance criteria are met.
[-]REP-005 — Drive and validate CAGE-2 replication across the CybORG sim and the ACES reference backend[/-][+]Deferred: CAGE-2 cross-backend replication (REP-005)[/+][-]Deferred: CAGE-2 cross-backend replication (REP-005)[/-][+]REP-005 — Drive and validate CAGE-2 replication across the CybORG sim and the ACES reference backend[/+]