FlyTris
MaleCNS mushroom body (KC→MBON, PAM / PPL1) learns Tetris as odor-gradient navigation with the fly’s own dopamine-gated plasticity rule.
Synapses change under a modeled rule grounded in fly biology — for example dopamine-gated change on Kenyon-cell → MBON synapses in the mushroom body. Learning begins here.
Dispute this rating →The evidence
No evidence, no level.The plastic weight matrix is the KC→MBON block of the MaleCNS connectome itself, extracted by neuron id rather than initialised at random.
https://github.com/rothilion26/flytris/blob/8574ddccfc0353e90d3758f62d2b3a3344fc80a6/flytris/sim/mushroom_body.py#L66-L75Those connectome weights are written to in place during play, by a dopamine-gated, compartment-specific rule driven by a reward-prediction error and an eligibility trace.
https://github.com/rothilion26/flytris/blob/8574ddccfc0353e90d3758f62d2b3a3344fc80a6/flytris/sim/plasticity.py#L145-L160A lesion table reports the control that matters: with the mushroom body silenced and the search policy untouched, learning does nothing at all (15.8 → 15.9 pieces, zero lines), so the gain comes from the circuit rather than the wrapper around it.
https://github.com/rothilion26/flytris/blob/8574ddccfc0353e90d3758f62d2b3a3344fc80a6/README.md
The clearest L2 in the index. The weight matrix that changes is built directly from MaleCNS KC→MBON edges, and the rule is dopamine-gated and compartment-specific. Not L3: the comparisons published are anatomical (Kenyon cell, APL and MBON counts against reported anatomy) and the body surrogate is calibrated against NeuroMechFly — another simulation, not measurements from real flies behaving.
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.