—
Project-defined temporal neural readout for candidate-training experiments.
CONNECTIVITY → ACTIVITY → BEHAVIOR
Observe fighting-game behavior and a fruit-fly body driven by neural activity.
A completed evaluation will appear here after publication.
Generation — · Not recorded · video — s
HP values describe the final recorded result, not the current playback frame.
Recorded evaluations and LIVE neural observations are separate records. Unrecorded activity is not invented for a video.
Actual MuJoCo physics, independent of FightingICE.
Body frames from another time are not overlaid on recordings.
NEURAL MOTOR DRIVE
Normalized command, 0–1. Not a physiological measure of muscle activity.
Body frames from another time are not overlaid on recordings.
NEURAL MOTOR DRIVE
Normalized command, 0–1. Not a physiological measure of muscle activity.
Project-defined temporal neural readout for candidate-training experiments.
Canonical and EMA-residual used the same checkpoint under held-out seed pairs.
Published candidates before this gate do not have a paired acceptance record.
Readout: Not recorded · mode Not recorded · policy_pixel_access=false
Current scheduled gate: utility — → — (—). Protocol: Not recorded.
Earlier paired-search pilot, parent: dealt — HP / took — HP / score —.
Proposal: dealt — HP / took — HP / score —.
Earlier pilot accepted update: —. Reward coefficients changed in that experiment: —.
Earlier paired-search evidence is not available.
Research checkpoint update
Latest completed fixed-condition evaluation
Candidates remain separate from approved inference
Generation counts and changed connections alone do not demonstrate learning effectiveness or improved performance.
Explicit evaluation and approval requiredfull-round protocol: max 60 s / 3600 frames · policy_pixel_access=false
Selected evaluation: Not recorded · seeds —/— · decision interval — frames
Reward ID: Not recorded · Model: Not recorded
State SHA-256: Not recorded
This viewer does not change reward settings. Evaluations with different protocols are not presented as matched comparisons.
Spectator data is separate from policy input.
Neural activity is simulated by applying the Shiu LIF model to MaleCNS connectivity. It is neither activity recorded from a living fly nor a whole-brain activity map. The viewer displays a bounded sample of top body IDs.
Game input/output and the MaleCNS-to-FlyBody actuator mapping are project-defined interfaces, not an experimentally identified muscle-innervation map. FightingICE coordinates do not position the fly.
LIVE neural activity and held body input are matched to observed decision history, and body PNG bytes are verified by SHA-256. FightingICE ScreenData is sampled independently; exact image-frame synchronization is not claimed. Physics can run slower than wall time under load.
FlyBody upstream: d015e9bfe441bd90ae431bac24c55cb74bdbce26
Adapter: malecns-annotated-motor-to-flybody-tripod-v2
Observation is read-only. Pixels are not policy inputs.
No verified LIVE snapshot is available in this view.
Source repository ↗