ProCreations/repro-why-linear-recurrent-memory-works-in-partially-observable-reinforcement-learning

Fix Hub short description length

Repair RingWorld comparison evidence

Verify RingWorld claim from exact source audit

claim 3: test Theorem 5.9's vanishing-error rate for the time-varying ALF in action-controlled POMDPs

Merge the eps*log(1/eps) rate verification into the canonical claim-2 page

Remove duplicate claim-2 page (merged into the canonical one)

Claim 2: fit the eps*log(1/eps) decoding-error rate against competing eps and eps^2 normalisations; log-form flat to 2.3x while alternatives spread 10x and 610x

Publish validated ICML reproduction for ywjHJIkUgW

initial commit