The experiment evaluates Rule-based reconfiguration under a transient byzantine / adversarial agents injection. Fault onset is randomized between environment steps 40 and 60.
Experimental trajectory graphs
One representative seed (seed 0 of 8) per target × mode combination from the full sweep. See the Results page for the multi-seed aggregate comparison across mechanisms. The graphs for every seed are archived at doi:10.5281/zenodo.23304785 (files available from 10 November 2026).
Rule-based reconfiguration
Expected evaluation record:
Metric
Recorded value
Degradation-area
5.37 ± 1.20 (corrected; mean ± 95% CI, n=72)
Safety-violation excess
21.7 ± 4.7 (weighted, control-subtracted; mean ± 95% CI, n=72)
Fraction of time healthy
0.697 ± 0.010 (mean ± 95% CI, n=72)
Relapse count
0.50 ± 0.12 (mean ± 95% CI, n=72)
False-recovery rate
0.000 ± 0.000 (mean ± 95% CI, n=72)
Intervention recall
1.000 ± 0.000 (mean ± 95% CI, n=72)
Values are means with 95% confidence intervals across every target × duration-fixed combination shown on this page (see graphs below), 8 seeds each. See the Results page for the cross-mechanism comparison.
Sorting Agent Target (fixed malicious mode)
Rule-based reconfiguration - Sorting Agent Target (fixed malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Pressing Agent Target (fixed malicious mode)
Rule-based reconfiguration - Pressing Agent Target (fixed malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Both Agents Target (fixed malicious mode)
Rule-based reconfiguration - Both Agents Target (fixed malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Sorting Agent Target (random malicious mode)
Rule-based reconfiguration - Sorting Agent Target (random malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Pressing Agent Target (random malicious mode)
Rule-based reconfiguration - Pressing Agent Target (random malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Both Agents Target (random malicious mode)
Rule-based reconfiguration - Both Agents Target (random malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Sorting Agent Target (worst action mode)
Rule-based reconfiguration - Sorting Agent Target (worst action mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Pressing Agent Target (worst action mode)
Rule-based reconfiguration - Pressing Agent Target (worst action mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Both Agents Target (worst action mode)
Rule-based reconfiguration - Both Agents Target (worst action mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Rule-based detect
Expected evaluation record:
Metric
Recorded value
Degradation-area
17.38 ± 3.38 (corrected; mean ± 95% CI, n=72)
Safety-violation excess
89.8 ± 18.8 (weighted, control-subtracted; mean ± 95% CI, n=72)
Fraction of time healthy
0.566 ± 0.023 (mean ± 95% CI, n=72)
Relapse count
0.69 ± 0.15 (mean ± 95% CI, n=72)
False-recovery rate
0.856 ± 0.020 (mean ± 95% CI, n=72)
Intervention recall
0.408 ± 0.051 (mean ± 95% CI, n=72)
Values are means with 95% confidence intervals across every target × duration-fixed combination shown on this page (see graphs below), 8 seeds each. See the Results page for the cross-mechanism comparison.
Sorting Agent Target (fixed malicious mode)
Rule-based detect - Sorting Agent Target (fixed malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Pressing Agent Target (fixed malicious mode)
Rule-based detect - Pressing Agent Target (fixed malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Both Agents Target (fixed malicious mode)
Rule-based detect - Both Agents Target (fixed malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Sorting Agent Target (random malicious mode)
Rule-based detect - Sorting Agent Target (random malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Pressing Agent Target (random malicious mode)
Rule-based detect - Pressing Agent Target (random malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Both Agents Target (random malicious mode)
Rule-based detect - Both Agents Target (random malicious mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Sorting Agent Target (worst action mode)
Rule-based detect - Sorting Agent Target (worst action mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Pressing Agent Target (worst action mode)
Rule-based detect - Pressing Agent Target (worst action mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Both Agents Target (worst action mode)
Rule-based detect - Both Agents Target (worst action mode) - evaluation trajectory under transient byzantine / adversarial agents (fault onset randomized between steps 40-60), seed 0 of the 8-seed sweep.
Notes
This page is intentionally separated from the paper-facing results so that full-resolution plots, additional seeds, and diagnostic figures can be documented without overloading the publication.
