A retrieval-augmented transformer with learned reference tokens shows marginal average MSE improvement over strong baselines on six time series benchmarks.
Quenched CLT for ancestral lineages of logistic branching random walks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We consider random walks in dynamic random environments which arise naturally as spatial embeddings of ancestral lineages in spatial locally regulated population models. In particular, as the main result, we prove the quenched central limit theorem for a random walk in dynamic random environment generated by time reversal of logistic branching random walks in a regime where the population density is sufficiently high. As an important tool we consider as auxiliary models random walks in dynamic random environments defined in terms of the time-reversal of oriented percolation. We show that the quenched central limit theorem holds if the influence of the random medium on the walks is suitably weak. The proofs of the quenched central limit theorems in these models rely on coarse-graining arguments and a construction of regeneration times for a pair of conditionally independent random walks in the same medium, combined with a coupling that relates them to a pair of independent random walks in two independent copies of the medium.
citation-role summary
background 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
TS-RAG: Retrieval Augmented Generation for Time Series Forecasting
A retrieval-augmented transformer with learned reference tokens shows marginal average MSE improvement over strong baselines on six time series benchmarks.