diff --git a/README.md b/README.md
index 7c4cd074..155acae4 100644
--- a/README.md
+++ b/README.md
@@ -8,6 +8,7 @@
 Miscellaneous functions for training and plotting classification and regression models.  Detailed documentation is at http://topepo.github.io/caret/index.html
 
 ## Install the current release from CRAN:
+
 ```r
 install.packages('caret')
 ```
@@ -16,3 +17,8 @@ install.packages('caret')
 ```r
 pak::pak('topepo/caret/pkg/caret')
 ```
+
+## Status
+
+caret will be 20 years old in March of 2026. The package is currently in *maintenance mode*; the author will fix bugs and make CRAN releases as needed, but there will not be any major features in the package. **It will stay on CRAN long-term**; it's not going away. 
+
diff --git a/pkg/caret/DESCRIPTION b/pkg/caret/DESCRIPTION
index 62fc9d86..20ea4b22 100644
--- a/pkg/caret/DESCRIPTION
+++ b/pkg/caret/DESCRIPTION
@@ -1,6 +1,6 @@
 Package: caret
 Title: Classification and Regression Training
-Version: 6.0-94
+Version: 7.0-1
 Authors@R: 
     c(person(given = "Max",
              family = "Kuhn",
diff --git a/pkg/caret/NAMESPACE b/pkg/caret/NAMESPACE
index b05d229a..5a4d42ac 100644
--- a/pkg/caret/NAMESPACE
+++ b/pkg/caret/NAMESPACE
@@ -38,11 +38,20 @@ S3method(dotplot,diff.resamples)
 S3method(dotplot,resamples)
 S3method(dummyVars,default)
 S3method(enumLC,default)
+S3method(enumLC,formula)
+S3method(enumLC,lm)
+S3method(enumLC,matrix)
 S3method(expoTrans,default)
+S3method(expoTrans,numeric)
 S3method(fitted,train)
 S3method(format,bagEarth)
 S3method(gafs,default)
 S3method(gafs,recipe)
+S3method(get_resample_perf,gafs)
+S3method(get_resample_perf,rfe)
+S3method(get_resample_perf,safs)
+S3method(get_resample_perf,sbf)
+S3method(get_resample_perf,train)
 S3method(ggplot,calibration)
 S3method(ggplot,gafs)
 S3method(ggplot,lift)
@@ -57,6 +66,7 @@ S3method(histogram,train)
 S3method(icr,default)
 S3method(icr,formula)
 S3method(knn3,data.frame)
+S3method(knn3,default)
 S3method(knn3,formula)
 S3method(knn3,matrix)
 S3method(knnreg,data.frame)
@@ -95,6 +105,7 @@ S3method(preProcess,default)
 S3method(precision,default)
 S3method(precision,table)
 S3method(predict,BoxCoxTrans)
+S3method(predict,PLS)
 S3method(predict,avNNet)
 S3method(predict,bag)
 S3method(predict,bagEarth)
@@ -131,6 +142,7 @@ S3method(print,avNNet)
 S3method(print,bag)
 S3method(print,bagEarth)
 S3method(print,bagFDA)
+S3method(print,bmars)
 S3method(print,calibration)
 S3method(print,classDist)
 S3method(print,confusionMatrix)
@@ -175,12 +187,14 @@ S3method(sbf,default)
 S3method(sbf,formula)
 S3method(sbf,recipe)
 S3method(sensitivity,default)
+S3method(sensitivity,matrix)
 S3method(sensitivity,table)
 S3method(sort,resamples)
 S3method(spatialSign,data.frame)
 S3method(spatialSign,default)
 S3method(spatialSign,matrix)
 S3method(specificity,default)
+S3method(specificity,matrix)
 S3method(specificity,table)
 S3method(splom,resamples)
 S3method(splsda,default)
diff --git a/pkg/caret/R/expoTrans.R b/pkg/caret/R/expoTrans.R
index b7c7a497..d0910e62 100644
--- a/pkg/caret/R/expoTrans.R
+++ b/pkg/caret/R/expoTrans.R
@@ -12,10 +12,10 @@ expoTrans.default <- function(y, na.rm  = TRUE, init = 0, lim = c(-4, 4), method
   rat <- max(y, na.rm = TRUE)/min(y, na.rm = TRUE)
   if(length(unique(y[!is.na(y)])) >= numUnique)
   {
-    results <- optim(init, manlyLik, x = y[!is.na(y)], neg = TRUE, method = method, 
-                     lower = lim[1], upper = lim[2]) 
+    results <- optim(init, manlyLik, x = y[!is.na(y)], neg = TRUE, method = method,
+                     lower = lim[1], upper = lim[2])
     out <- list(lambda = results$par, est = manly(y, results$par))
-    if(length(unique(out$est)) == 1 | results$convergence > 0) 
+    if(length(unique(out$est)) == 1 | results$convergence > 0)
       out <- list(lambda = NA, est = y)
   } else out <- list(lambda = NA, est = y)
   out$n <- sum(!is.na(y))
@@ -27,7 +27,7 @@ expoTrans.default <- function(y, na.rm  = TRUE, init = 0, lim = c(-4, 4), method
   out
 }
 
-#' @importFrom stats optim
+#' @export
 expoTrans.numeric <- function(y, na.rm  = TRUE, init = 0, lim = c(-4, 4), method = "Brent", numUnique = 3, ...)
 {
   requireNamespaceQuietStop("e1071")
@@ -36,10 +36,10 @@ expoTrans.numeric <- function(y, na.rm  = TRUE, init = 0, lim = c(-4, 4), method
   rat <- max(y, na.rm = TRUE)/min(y, na.rm = TRUE)
   if(length(unique(y[!is.na(y)])) >= numUnique)
   {
-    results <- optim(init, manlyLik, x = y[!is.na(y)], neg = TRUE, method = method, 
-                     lower = lim[1], upper = lim[2]) 
+    results <- optim(init, manlyLik, x = y[!is.na(y)], neg = TRUE, method = method,
+                     lower = lim[1], upper = lim[2])
     out <- list(lambda = results$par, est = manly(y, results$par))
-    if(length(unique(out$est)) == 1 | results$convergence > 0) 
+    if(length(unique(out$est)) == 1 | results$convergence > 0)
       out <- list(lambda = NA, est = y)
   } else out <- list(lambda = NA, est = y)
   out$n <- sum(!is.na(y))
@@ -52,15 +52,15 @@ expoTrans.numeric <- function(y, na.rm  = TRUE, init = 0, lim = c(-4, 4), method
 }
 
 #' @export
-print.expoTrans <- function(x, digits = max(3L, getOption("digits") - 3L), ...) 
+print.expoTrans <- function(x, digits = max(3L, getOption("digits") - 3L), ...)
 {
   cat("Exponential Transformation\n\n")
-  
+
   cat(x$n, "data points used to estimate Lambda\n\n")
   cat("Input data summary:\n")
   print(x$summary)
   cat("\nLargest/Smallest:", signif(x$ratio, digits), "\n")
-  cat("Sample Skewness:", signif(x$skewness, digits), "\n")      
+  cat("Sample Skewness:", signif(x$skewness, digits), "\n")
   if(!is.na(x$lambda))
   {
     cat("\nEstimated Lambda:", signif(x$lambda, digits), "\n")
@@ -68,7 +68,7 @@ print.expoTrans <- function(x, digits = max(3L, getOption("digits") - 3L), ...)
   cat("\n")
   invisible(x)
 }
-  
+
 #' @export
 predict.expoTrans <- function(object, newdata, ...)
 {
@@ -80,8 +80,8 @@ predict.expoTrans <- function(object, newdata, ...)
   out
 }
 
-  
-manly <- function(x, lambda) 
+
+manly <- function(x, lambda)
   if(lambda == 0) x else (exp(lambda*x) - 1)/lambda
 
 #' @importFrom stats var
@@ -93,7 +93,7 @@ manlyLik <- function(lambda, x, neg = FALSE)
   L2 <- .5 * sum(y - mean(y, na.rm = TRUE))/v
   L3 <- sum(!is.na(x)) * log(sqrt(2*pi)*sqrt(v))
   out <- L1 - L2 - L3
-  if(!is.finite(out) | is.na(out)) out <- .Machine$double.xmax  
+  if(!is.finite(out) | is.na(out)) out <- .Machine$double.xmax
   if(neg) -out else out
 }
 
diff --git a/pkg/caret/R/findCorrelation.R b/pkg/caret/R/findCorrelation.R
index f6b98b9e..d1ccc2c4 100644
--- a/pkg/caret/R/findCorrelation.R
+++ b/pkg/caret/R/findCorrelation.R
@@ -119,9 +119,7 @@ findCorrelation_exact <- function(x, cutoff = 0.90, verbose = FALSE)
 #' = TRUE}) otherwise a vector of column names. If no correlations meet the
 #' criteria, \code{integer(0)} is returned.
 #' @author Original R code by Dong Li, modified by Max Kuhn
-#' @seealso \code{\link[subselect:eleaps]{leaps}},
-#' \code{\link[subselect:genetic]{genetic}},
-#' \code{\link[subselect:anneal]{anneal}}, \code{\link{findLinearCombos}}
+#' @seealso \code{\link{findLinearCombos}}
 #' @keywords manip
 #' @examples
 #'
diff --git a/pkg/caret/R/findLinearCombos.R b/pkg/caret/R/findLinearCombos.R
index fa3c9b00..b09df70e 100644
--- a/pkg/caret/R/findLinearCombos.R
+++ b/pkg/caret/R/findLinearCombos.R
@@ -9,12 +9,14 @@ enumLC.default <- function(object, ...)
    stop(paste('enumLC does not support ', class(object), 'objects'))
 }
 
+#' @export
 enumLC.matrix <- function(object, ...)
 {
    # factor the matrix using QR decomposition and then process it
    internalEnumLC(qr(object))
 }
 
+#' @export
 enumLC.lm <- function(object, ...)
 {
    # extract the QR decomposition and the process it
@@ -22,6 +24,7 @@ enumLC.lm <- function(object, ...)
 }
 
 #' @importFrom stats lm
+#' @export
 enumLC.formula <- function(object, ...)
 {
    # create an lm fit object from the formula, and then call
@@ -74,7 +77,6 @@ internalEnumLC <- function(qrObj, ...)
 #' numbers that can be removed to counter the linear combinations}
 #' @author Kirk Mettler and Jed Wing (\code{enumLC}) and Max Kuhn
 #' (\code{findLinearCombos})
-#' @seealso \code{\link[subselect:trim.matrix]{trim.matrix}}
 #' @keywords manip
 #' @examples
 #'
diff --git a/pkg/caret/R/knn3.R b/pkg/caret/R/knn3.R
index 68f135e8..c08f21d0 100644
--- a/pkg/caret/R/knn3.R
+++ b/pkg/caret/R/knn3.R
@@ -54,9 +54,10 @@
 #' cl <- factor(c(rep("s",25), rep("c",25), rep("v",25)))
 #' knn3Train(train, test, cl, k = 5, prob = TRUE)
 #'
-#' @export knn3
+#' @export
 "knn3" <- function(x, ...)   UseMethod("knn3")
 
+#' @export
 knn3.default <- function(x, ...)
 {
    if(!inherits(x, "formula"))  stop("knn3 only implemented for formula objects")
diff --git a/pkg/caret/R/misc.R b/pkg/caret/R/misc.R
index 6e5c1f22..c30c4bd4 100644
--- a/pkg/caret/R/misc.R
+++ b/pkg/caret/R/misc.R
@@ -337,39 +337,42 @@ requireNamespaceQuietStop <- function(package) {
 
 get_resample_perf <- function (x, ...) UseMethod("get_resample_perf")
 
-get_resample_perf.train <- function(x) {
+#' @export
+get_resample_perf.train <- function(x, ...) {
   if(x$control$returnResamp == "none")
     stop("use returnResamp == 'none' in trainControl()", call. = FALSE)
   out <- merge(x$resample, x$bestTune)
   out[, c(x$perfNames, "Resample")]
 }
 
-get_resample_perf.rfe <- function(x) {
+#' @export
+get_resample_perf.rfe <- function(x, ...) {
   if(x$control$returnResamp == "none")
     stop("use returnResamp == 'none' in trainControl()", call. = FALSE)
   out <- subset(x$resample, Variables == x$bestSubset)
   out[, c(x$perfNames, "Resample")]
 }
 
-get_resample_perf.sbf <- function(x) {
+#' @export
+get_resample_perf.sbf <- function(x, ...) {
   if(x$control$returnResamp == "none")
     stop("use returnResamp == 'none' in trainControl()", call. = FALSE)
   x$resample
 }
 
-get_resample_perf.safs <- function(x) {
+#' @export
+get_resample_perf.safs <- function(x, ...) {
   out <- subset(x$external, Iter == x$optIter)
   out[, !(names(out) %in% "Iter")]
 }
 
-get_resample_perf.gafs <- function(x) {
+#' @export
+get_resample_perf.gafs <- function(x, ...) {
   out <- subset(x$external, Iter == x$optIter)
   out[, !(names(out) %in% "Iter")]
 }
 
 
-
-
 #' Sequences of Variables for Tuning
 #'
 #' This function generates a sequence of \code{mtry} values for random forests.
diff --git a/pkg/caret/R/predict.PLS.R b/pkg/caret/R/predict.PLS.R
index e3cc4951..0209c620 100644
--- a/pkg/caret/R/predict.PLS.R
+++ b/pkg/caret/R/predict.PLS.R
@@ -1,5 +1,6 @@
 
-predict.PLS <- function(object, newdata, 
+#' @export
+predict.PLS <- function(object, newdata,
    ncomp = NULL,
    type = ifelse(object$isRegression, "response", "class"), ...)
 {
@@ -7,7 +8,7 @@ predict.PLS <- function(object, newdata,
    if(is.null(ncomp) & !is.null(object$bestIter$.ncomp)) ncomp <- object$bestIter$.ncomp
 
    if(is.null(ncomp)) stop("the number of components must be given")
-   
+
    # adapted from the pls package
    if(object$isRegression & c(type %in% c("class", "prob")))
       stop("cannot get a class estimate if the original y was not a factor")
@@ -29,25 +30,25 @@ predict.PLS <- function(object, newdata,
    for (i in seq(along.with = 1:ncomp)) BInt[1, , i] <- object$Ymeans - object$Xmeans %*% B[, , i]
    B <- BInt
    # stop
-   
-   # from predict.mvr in pls package   
+
+   # from predict.mvr in pls package
    dPred <- dim(B)
    dPred[1] <- dim(newdata)[1]
    dnPred <- dimnames(B)
    dnPred[1] <- dimnames(newdata)[1]
    pred <- array(dim = dPred, dimnames = dnPred)
-   predY <- sweep(newdata %*% B[-1, , ncomp], 2, B[1, , ncomp], "+")   
+   predY <- sweep(newdata %*% B[-1, , ncomp], 2, B[1, , ncomp], "+")
    # stop
- 
+
    out <- switch(
       type,
       response = predY,
-      class = 
+      class =
          {
             classNum <- apply(predY, 1, which.max)
-            factor(object$yLevels[classNum], levels = object$yLevels)   
+            factor(object$yLevels[classNum], levels = object$yLevels)
          },
-      # use softmax technique here         
+      # use softmax technique here
       prob =  t(apply(predY, 1, function(data) exp(data)/sum(exp(data)))))
 
    out
diff --git a/pkg/caret/R/print.mars.R b/pkg/caret/R/print.mars.R
index c125711f..9323b06a 100644
--- a/pkg/caret/R/print.mars.R
+++ b/pkg/caret/R/print.mars.R
@@ -1,4 +1,5 @@
-print.bmars <- function (x, ...) 
+#' @export
+print.bmars <- function (x, ...)
 {
     cat("\nCall:\n", deparse(x$call), "\n\n", sep = "")
     if(!is.null(x$x))cat("Data:\n   # variables:\t", dim(x$x)[2], "\n   # samples:\t", dim(x$x)[1], "\n")
@@ -6,10 +7,10 @@ print.bmars <- function (x, ...)
       "\nModel:",
       "\n   B:        \t", x$B,
       "\n   degree:   \t", x$fit[[1]]$degree,
-      "\n   nk:       \t", x$fit[[1]]$nk,      
-      "\n   penalty:  \t", x$fit[[1]]$penalty,      
+      "\n   nk:       \t", x$fit[[1]]$nk,
+      "\n   penalty:  \t", x$fit[[1]]$penalty,
       "\n   threshold:\t", x$fit[[1]]$thresh, "\n")
-     
+
     cat("\n")
     invisible(x)
 }
diff --git a/pkg/caret/R/sensitivity.R b/pkg/caret/R/sensitivity.R
index 7c68d383..7c3dbd2c 100644
--- a/pkg/caret/R/sensitivity.R
+++ b/pkg/caret/R/sensitivity.R
@@ -196,6 +196,7 @@ sensitivity <-
 }
 
 #' @rdname sensitivity
+#' @export
 "sensitivity.matrix" <-
   function(data, positive = rownames(data)[1], ...)
 {
diff --git a/pkg/caret/R/specificity.R b/pkg/caret/R/specificity.R
index ae39d995..07b83bdd 100644
--- a/pkg/caret/R/specificity.R
+++ b/pkg/caret/R/specificity.R
@@ -65,6 +65,7 @@ function(data, reference, negative = levels(reference)[-1], na.rm = TRUE, ...)
   spec
 }
 
+#' @export
 "specificity.matrix" <-
   function(data, negative = rownames(data)[-1], ...)
 {
diff --git a/pkg/caret/R/train.default.R b/pkg/caret/R/train.default.R
index 7c3ca134..9263d224 100644
--- a/pkg/caret/R/train.default.R
+++ b/pkg/caret/R/train.default.R
@@ -47,7 +47,7 @@
 #'  must have column names (see Details below). Preprocessing using
 #'  the \code{preProcess} argument only supports matrices or data
 #'  frames. When using the recipe method, \code{x} should be an
-#'  unprepared \code{\link{recipe}} object that describes the model
+#'  unprepared [recipes::recipe()] object that describes the model
 #'  terms (i.e. outcome, predictors, etc.) as well as any
 #'  pre-processing that should be done to the data. This is an
 #'  alternative approach to specifying the model. Note that, when
diff --git a/pkg/caret/inst/NEWS.Rd b/pkg/caret/inst/NEWS.Rd
index 9420a82a..f7f58cf5 100644
--- a/pkg/caret/inst/NEWS.Rd
+++ b/pkg/caret/inst/NEWS.Rd
@@ -3,6 +3,13 @@
 \newcommand{\cpkg}{\href{https://CRAN.R-project.org/package=#1}{\pkg{#1}}}
 \newcommand{\issue}{\href{https://github.com/topepo/caret/issues/#1}{(issue #1)}}
 
+\section{Changes in version 7.0-1}{
+ \itemize{
+   \item CRAN mandated update.
+   \item caret will be 20 years old in March of 2026. The package is currently in maintenance mode; the author will fix bugs and make CRAN releases as needed, but there will not be any major features in the package. It will stay on CRAN long-term; it's not going away.
+  }
+}
+
 \section{Changes in version 6.0-94}{
   \itemize{
    \item Bug fix in how some S3 signatures were designed (for R-devel).
@@ -192,7 +199,7 @@
     \item Switched to non-formula interface in ranger. Also, another tuning parameter was added to ranger (\code{splitrule}) that can be used to change the splitting procedure and includes extremely randomized trees. This requires version 0.8.0 of the \cpkg{ranger} package. \issue{581}
     \item A simple "null model" was added. For classification, it predictors using the most prevalent level and, for regression, fits an intercept only model. \issue{694}
     \item A function \code{thresholder} was added to analyze the resample results for two class problems to choose an appropriate probability cutoff a la \url{https://topepo.github.io/caret//using-your-own-model-in-train.html#Illustration5} \issue{224}.
-    \item Two neural network models (containing a single hidden layers) using \code{tensorflow}/\code{keras} were added.  \code{mlpKerasDecay} uses standard weight decay while \code{mlpKerasDropout} uses dropout for regularization. Both use RMSProp optimizer and have a lot of tuning parameters. Two additional models, \code{mlpKerasDecayCost} and \code{mlpKerasDropoutCost}, are classification only and perform cost-sensitive learning. Note that these models will not run in parallel using \cpkg{caret}'s parallelism and also will not give reproducible results from run-to-run (see \url{https://github.com/rstudio/keras/issues/42}).
+    \item Two neural network models (containing a single hidden layers) using \code{tensorflow}/\code{keras} were added.  \code{mlpKerasDecay} uses standard weight decay while \code{mlpKerasDropout} uses dropout for regularization. Both use RMSProp optimizer and have a lot of tuning parameters. Two additional models, \code{mlpKerasDecayCost} and \code{mlpKerasDropoutCost}, are classification only and perform cost-sensitive learning. Note that these models will not run in parallel using \cpkg{caret}'s parallelism and also will not give reproducible results from run-to-run (see \url{https://github.com/rstudio/keras3/issues/42}).
     \item The range for one parameter (\code{gamma}) was modified in the \code{mlpSGD} model code.
     \item A bug in classification models with all missing predictions was fixed (found by andzandz11). \issue{684}
     \item A bug preventing preprocessing to work properly when the preprocessing transformations are related to individual columns only fixed by Mateusz Kobos in \issue{679}.
diff --git a/pkg/caret/man/findCorrelation.Rd b/pkg/caret/man/findCorrelation.Rd
index c4ae2675..d847ba5d 100644
--- a/pkg/caret/man/findCorrelation.Rd
+++ b/pkg/caret/man/findCorrelation.Rd
@@ -76,9 +76,7 @@ findCorrelation(R2, cutoff = .99, verbose = TRUE)
 
 }
 \seealso{
-\code{\link[subselect:eleaps]{leaps}},
-\code{\link[subselect:genetic]{genetic}},
-\code{\link[subselect:anneal]{anneal}}, \code{\link{findLinearCombos}}
+\code{\link{findLinearCombos}}
 }
 \author{
 Original R code by Dong Li, modified by Max Kuhn
diff --git a/pkg/caret/man/findLinearCombos.Rd b/pkg/caret/man/findLinearCombos.Rd
index 27c79c06..a1c0d62c 100644
--- a/pkg/caret/man/findLinearCombos.Rd
+++ b/pkg/caret/man/findLinearCombos.Rd
@@ -49,9 +49,6 @@ testData2[,6] <- c(0, 0, 1, 0, 0, 1)
 
 findLinearCombos(testData2)
 
-}
-\seealso{
-\code{\link[subselect:trim.matrix]{trim.matrix}}
 }
 \author{
 Kirk Mettler and Jed Wing (\code{enumLC}) and Max Kuhn
diff --git a/pkg/caret/man/train.Rd b/pkg/caret/man/train.Rd
index 45dc771e..0e862d36 100644
--- a/pkg/caret/man/train.Rd
+++ b/pkg/caret/man/train.Rd
@@ -44,7 +44,7 @@ simple matrix, data frame or other type (e.g. sparse matrix) but
 must have column names (see Details below). Preprocessing using
 the \code{preProcess} argument only supports matrices or data
 frames. When using the recipe method, \code{x} should be an
-unprepared \code{\link{recipe}} object that describes the model
+unprepared [recipes::recipe()] object that describes the model
 terms (i.e. outcome, predictors, etc.) as well as any
 pre-processing that should be done to the data. This is an
 alternative approach to specifying the model. Note that, when