Demo video
Baseline NPC is exploited by scripted kiting. Psydust NPC resists under the same benchmark scenario.
Scenario
The benchmark uses a fixed Unity scene where a target/player kites a Psydust-controlled chaser. The run records metrics, inputs, outputs, and replay configuration for the frozen scenario.
The scenario is game-based, but the verified properties here — trace generation, output-bounds checking, and deterministic replay — are the domain-neutral parts of the Psydust system.
Status
Psydust Verification Level: P3 - Replay Determinism Verified
Internal Psydust benchmark report, not third-party certification.
Run 003 Summary
f40a502e77850e3073733232798099f6656f70facc025990c038018670f8e841
What this proves
Benchmark evidence was generated for the frozen kite-resistance scenario.
Psydust outputs stayed within the declared bounds for the run.
Recorded outputs replayed exactly from recorded inputs.
Scope and limitations
Current scope:
This benchmark verifies one frozen Unity kite-resistance run, including metrics, trace presence, output bounds, and deterministic replay.
Limitations:
This is an internal Psydust benchmark report, not third-party certification or production-wide validation.
Unity Reference Integration v0.1
A minimal Unity integration showing how a Psydust behavior module receives game-state inputs, produces bounded outputs, and drives NPC movement decisions inside the Kite Resistance Benchmark scene.
This is a reference integration, not a full SDK.
Private Evaluation Package
Send one enemy behavior and one exploit pattern. Psydust produces a before/after benchmark run and Trust Report.
Inputs:
- one enemy behavior
- one exploit pattern, such as kiting, doorway farming, leash abuse, or safe-zone farming
Outputs:
- baseline vs Psydust result
- metrics
- replay-verified Trust Report
- optional Unity behavior module for evaluation