Spaces:

---

LimmeDev

/

manifold-cs2-training

like 1

Running on Zero

MANIFOLD (Motor-Aware Neural Inference for Faithfulness Of Latent Dynamics) is a custom architecture for CS2 cheat detection. It models player behavior as latent dynamics and uses motor/physics constraints to better separate legitimate skill from mechanically implausible patterns.

A key property of MANIFOLD is adaptive detectability: as cheating effectiveness rises (e.g., unusually high K/D, headshot rate, and related signals), the model’s confidence and detection strength are designed to scale accordingly, with uncertainty quantification to avoid overconfident false positives.