both fixes applied

two fixes: separation achieved

autoresearch: silicon directions + community proposal

Multi-step + Loihi energy: honest ~4% energy restatement

Multi-step + Loihi energy: honest ~4% energy restatement

Multi-step + Loihi energy: honest ~4% energy restatement

SNN Exp1+2: TTFS + spike-accurate energy

SNN Exp1+2: TTFS + spike-accurate energy

SNN Exp1+2: TTFS + spike-accurate energy

Method paper: add SNN spike-timing results section

Method paper: add SNN spike-timing results section

CVaR-DQN evolution: spike-timing risk coding (Martinet-lens)

CVaR-DQN evolution: spike-timing risk coding (Martinet-lens)

CVaR-DQN evolution: spike-timing risk coding (Martinet-lens)

Add method-focused CVaR-DQN paper (Nachabe & Callens)

Add method-focused CVaR-DQN paper (Nachabe & Callens)

v10I XL head-to-head: paper §8.2 + results (1000-TU graded-hazard)

v10I XL head-to-head: paper §8.2 + results (1000-TU graded-hazard)

v10I XL head-to-head: paper §8.2 + results (1000-TU graded-hazard)

Add v10H arXiv paper (PDF+tex): 3 peer-review loops, authors, Amadeus/MLGO context

Add v10H arXiv paper (PDF+tex): 3 peer-review loops, authors, Amadeus/MLGO context

Add best CVaR-DQN model (seed 789, alpha=0.6) for the head-to-head

v10H FINAL: CVaR-DQN head-to-head verdict (null result, §8.1)

v10H paper: CVaR-DQN implementation (§6.1) + live head-to-head (§8.1)

v10g2: correctness-priced DQN (positive result: 1.174x, 0% regression)

initial commit