flydoom
Alternate Doom fly on Princeton’s FAFB-v783 (FlyWire) dataset rather than MaleCNS.
Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.
Dispute this rating →The evidence
No evidence, no level.The simulator implements a documented LIF model with membrane and synaptic time constants, threshold, reset and refractory period.
https://github.com/mutkuoz/flydoom/blob/dffc37fc48f7736d6614f5bc240694971900c824/flydoom/lif.py#L1-L21Short-term depression is off by default and the reported model is explicitly the frozen one.
https://github.com/mutkuoz/flydoom/blob/dffc37fc48f7736d6614f5bc240694971900c824/flydoom/config.py#L574-L575The project states that its dopaminergic teaching signal reaches a brain with no plasticity to receive it.
https://github.com/mutkuoz/flydoom/blob/dffc37fc48f7736d6614f5bc240694971900c824/flydoom/registry.py#L406-L414
LIF simulation of all 139,255 FAFB-v783 neurons in the model form of Shiu et al. (tau_mem, tau_syn, threshold, reset, refractory, conduction delay), stepped at 0.5 ms. Not L2: nothing writes a synaptic weight during a run. The one plasticity mechanism in the codebase is short-term depression, and it is (a) disabled by default, with the config stating 'The reported model is the frozen one', and (b) a per-neuron presynaptic release resource rather than a per-synapse weight, so even when enabled it is vesicle depletion, not learning. The project's own neuron registry says the punishment-dopamine pathway 'injects into a brain with no plasticity to receive it'. This is a third, independent flydoom rated on its own source; it also reports its own negative result on motion vision in detail, which is good science and is not held against it.
Record
Nearby
All projects →MaleCNS simulation connected to ViZDoom through modeled visual inputs and a fixed button readout. Includes plasticity experiments and negative validation results; learned survival has not been demonstrated.
Research framework comparing a MaleCNS-constrained sparse recurrent controller with rewired graphs, conventional neural networks, and other controls on simple VizDoom tasks.
Ratings on thelearningfly.com are proposals, not verdicts. Every one of them is arguable in public.