Small Recurrent Network
A specialized policy network trained inside a simulator that processes structured game-state features and outputs voting decisions. It does not handle natural language, and is repeatedly calibrated through large numbers of matches.
Timeline (last 90 days)
The small recurrent network in the experiment is a policy network trained inside the simulator, taking structured game-state features as input and outputting a suspicion score for each player, does not process natural language, and derives all its advantage from repeated calibration over thousands of rounds.
When the experiment added the voting network's intuition to the crew prompt, the intuition line stated that the model can pick the real impostor about 70% of the time, but that 70% was a rough estimate the author made when writing the prompt, not an actual measurement.
All Facts (2)
The small recurrent network in the experiment is a policy network trained inside the simulator, taking structured game-state features as input and outputting a suspicion score for each player, does not process natural language, and derives all its advantage from repeated calibration over thousands of rounds.
50%UnverifiedWhen the experiment added the voting network's intuition to the crew prompt, the intuition line stated that the model can pick the real impostor about 70% of the time, but that 70% was a rough estimate the author made when writing the prompt, not an actual measurement.
50%