Viz

Visualizations from reinforcement learning, large-scale training, and simulation. Extracted from research notes and dev logs.

Cable Mind: UR5e Demo UR5e + Robotiq 2F-85, vision pipeline
Cable Mind: 4-Camera Montage
Cable Mind: 4-Camera Montage Hybrid camera placement for triangulated views
The Abyss (2,500 Agents) 500 platforms, single shared policy
MARL Simulation (Unreal) Multi-agent RL in Unreal Engine
Amax Explosion @ 1.7B
Amax Explosion @ 1.7B HC vs mHC signal amplification
Layer-wise Amax Heatmap
Layer-wise Amax Heatmap Signal amplification by layer
Instability Scaling Law
Instability Scaling Law Amax vs model depth

Interactive

Signal Flow: Residual vs Hyper-Connection

Standard residuals use one stream. Hyper-Connections use n parallel streams with learnable mixing matrices (H_res). Full explanation →

Standard Residual F Hyper-Connection H_res H_pre H_post F

Amax Counter

HC hits 10,924x signal amplification at 1.7B parameters. mHC stays at 1.0. Full explanation →

HC
1.00
starting
mHC
1.00
stable
Step: 0 / 5,000

Layer Heatmap

Instability starts at Layer 0, the input embedding. Not a deep network problem. Full explanation →

HC
mHC
1.0 1.5 2.0+
Step: 0 / 5,000

Sinkhorn-Knopp

Alternating row/column normalization converges to doubly stochastic. The fix. Details →

0.40 0.20 0.30 0.10 0.20 0.30 0.20 0.30 0.30 0.20 0.40 0.10 0.10 0.30 0.10 0.50 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 row col
Converged
Iteration 5/5