RL Agent Status

Live status for the DQN agents used in signal generation.

Q Equity DQN

Loading Q status...

Atlas Crypto DQN

Loading Atlas status...

How to interpret:

Training Steps: total learning updates. More steps usually means more stable behavior.
Epsilon: exploration rate. Higher means more random actions; lower means more exploitation.
Stage: training-volume label, not a guarantee of future returns.
Runtime: should show DQN when neural weights are loaded.
Checkpoint Format: storage schema version only, not the number of past model revisions.
Cloud Enabled: Yes means RL state persists across restarts and users.
Universe scans: scanning does not train the model live. Training steps move when learning updates are saved from evaluated outcomes and backtests.

Q is trained on equity data. Atlas is trained on crypto data (BTC, ETH, SOL). They are completely separate DQN instances — no weight sharing.
Both agents are one vote inside a multi-module decision system, not standalone trading systems.