Cohesion

Does your agent's reasoning hold together?

Your own 0G API key (optional)

Every run below spends real 0G inference calls. Leave this blank to use this deployment's key; fill it in and your calibrations/checks run against your own 0G balance instead — it's sent as a request header, never stored server-side.

The three propositions

A = WETH/USDC ends higher in 24h

B = USDC/WBTC ends higher in 24h

C = WBTC/WETH ends higher in 24h

The agent is asked about them in three separate, isolated calls, e.g. context (A,B) is only asked about A and B, blind to the third proposition. This forms partial probability distributions.

After that we're using an expectation value for a full (A,B,C) distribution, which should equal exactly 2.0 ideally, as a measure of the coherence of the model the agent is using.

Calibrate an agent's health baseline Model (subject under test)

Only models with a healthy provider on 0G's router are listed.

System prompt (optional) Trading pair Repetitions per context: 12

Trade — advisory coherence gate Model (subject under test)

Only models with a stored baseline are listed — calibrate a model above before it can be checked.

Trading pair Amount in Repetitions per context: 3

Stored baselines

Loading…