jgrusewski 98afee0a51 plan: 9-exposure action space + dual C51/IQN distributional architecture ...

Phase A: expand exposure 5→9 (25% steps, 81 factored actions)
Phase B: wire existing GpuIqnHead for CVaR position scaling —
C51 picks direction, IQN sizes the risk.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-03-23 09:03:19 +01:00

..

plans

plan: 9-exposure action space + dual C51/IQN distributional architecture

2026-03-23 09:03:19 +01:00

specs

spec: 9-exposure action space design — expand from 5 to 9 levels

2026-03-23 08:45:18 +01:00