diff --git a/python/doc/source/reference/functions.metrics.rst b/python/doc/source/reference/functions.metrics.rst
index b7e046a93..1a0f102c0 100644
--- a/python/doc/source/reference/functions.metrics.rst
+++ b/python/doc/source/reference/functions.metrics.rst
@@ -106,13 +106,13 @@ is defined as :cite:p:`ZitThiLauFon2003:tec`
 .. math::
    epsilon(A,R) = \max_{r \in R} \min_{a \in A} \max_{1 \leq i \leq m} epsilon(a_i, r_i)
 
-where :math:`a` and :math:`b` are objective vectors of length :math:`m`.
+where :math:`a` and :math:`r` are objective vectors of length :math:`m`.
 
-In the case of minimization of objective :math:`i`, :math:`epsilon(a_i,b_i)` is
-computed as :math:`a_i/b_i` for the multiplicative variant (respectively,
-:math:`a_i - b_i` for the additive variant), whereas in the case of
-maximization of objective :math:`i`, :math:`epsilon(a_i,b_i) = b_i/a_i` for the
-multiplicative variant (respectively, :math:`b_i - a_i` for the additive
+In the case of minimization of objective :math:`i`, :math:`epsilon(a_i,r_i)` is
+computed as :math:`a_i/r_i` for the multiplicative variant (respectively,
+:math:`a_i - r_i` for the additive variant), whereas in the case of
+maximization of objective :math:`i`, :math:`epsilon(a_i,r_i) = r_i/a_i` for the
+multiplicative variant (respectively, :math:`r_i - a_i` for the additive
 variant).  This allows computing a single value for problems where some
 objectives are to be maximized while others are to be minimized. Moreover, a
 lower value corresponds to a better approximation set, independently of the
diff --git a/r/R/epsilon.R b/r/R/epsilon.R
index 5657326f5..4d2942150 100644
--- a/r/R/epsilon.R
+++ b/r/R/epsilon.R
@@ -22,11 +22,11 @@
 #'
 #' where \eqn{a} and \eqn{b} are objective vectors of length \eqn{m}.
 #'
-#' In the case of minimization of objective \eqn{i}, \eqn{epsilon(a_i,b_i)} is
-#' computed as \eqn{a_i/b_i} for the multiplicative variant (respectively,
-#' \eqn{a_i - b_i} for the additive variant), whereas in the case of
-#' maximization of objective \eqn{i}, \eqn{epsilon(a_i,b_i) = b_i/a_i} for the
-#' multiplicative variant (respectively, \eqn{b_i - a_i} for the additive
+#' In the case of minimization of objective \eqn{i}, \eqn{epsilon(a_i,r_i)} is
+#' computed as \eqn{a_i/r_i} for the multiplicative variant (respectively,
+#' \eqn{a_i - r_i} for the additive variant), whereas in the case of
+#' maximization of objective \eqn{i}, \eqn{epsilon(a_i,r_i) = r_i/a_i} for the
+#' multiplicative variant (respectively, \eqn{b_i - r_i} for the additive
 #' variant). This allows computing a single value for problems where some
 #' objectives are to be maximized while others are to be minimized. Moreover, a
 #' lower value corresponds to a better approximation set, independently of the
diff --git a/r/man/epsilon.Rd b/r/man/epsilon.Rd
index 9bff5842d..5a8309a73 100644
--- a/r/man/epsilon.Rd
+++ b/r/man/epsilon.Rd
@@ -33,13 +33,13 @@ reference set \eqn{R \subset \mathbb{R}^m} is defined as
 
 \deqn{epsilon(A,R) = \max_{r \in R} \min_{a \in A} \max_{1 \leq i \leq m} epsilon(a_i, r_i)}
 
-where \eqn{a} and \eqn{b} are objective vectors of length \eqn{m}.
+where \eqn{a} and \eqn{r} are objective vectors of length \eqn{m}.
 
-In the case of minimization of objective \eqn{i}, \eqn{epsilon(a_i,b_i)} is
-computed as \eqn{a_i/b_i} for the multiplicative variant (respectively,
-\eqn{a_i - b_i} for the additive variant), whereas in the case of
-maximization of objective \eqn{i}, \eqn{epsilon(a_i,b_i) = b_i/a_i} for the
-multiplicative variant (respectively, \eqn{b_i - a_i} for the additive
+In the case of minimization of objective \eqn{i}, \eqn{epsilon(a_i,r_i)} is
+computed as \eqn{a_i/r_i} for the multiplicative variant (respectively,
+\eqn{a_i - r_i} for the additive variant), whereas in the case of
+maximization of objective \eqn{i}, \eqn{epsilon(a_i,r_i) = r_i/a_i} for the
+multiplicative variant (respectively, \eqn{r_i - a_i} for the additive
 variant). This allows computing a single value for problems where some
 objectives are to be maximized while others are to be minimized. Moreover, a
 lower value corresponds to a better approximation set, independently of the