diff --git a/source/isaaclab_tasks/changelog.d/mataylor-fix-tqdm-ci-logs.rst b/source/isaaclab_tasks/changelog.d/mataylor-fix-tqdm-ci-logs.rst
new file mode 100644
index 000000000000..a67a9ed5fcdd
--- /dev/null
+++ b/source/isaaclab_tasks/changelog.d/mataylor-fix-tqdm-ci-logs.rst
@@ -0,0 +1,8 @@
+Fixed
+^^^^^
+
+* Fixed the "Prefilling reset buffer" progress bar in
+  :func:`~isaaclab_tasks.core.lift.mdp.events` rendering as garbled block
+  characters in GitHub Actions logs. The :class:`tqdm` bar is now disabled when
+  ``stderr`` is not a TTY; a plain :mod:`logging` message is emitted at the
+  start and on completion so CI logs still show meaningful progress.
diff --git a/source/isaaclab_tasks/isaaclab_tasks/core/lift/mdp/events.py b/source/isaaclab_tasks/isaaclab_tasks/core/lift/mdp/events.py
index cc4f86a2a3ef..6822301dd724 100644
--- a/source/isaaclab_tasks/isaaclab_tasks/core/lift/mdp/events.py
+++ b/source/isaaclab_tasks/isaaclab_tasks/core/lift/mdp/events.py
@@ -7,6 +7,8 @@
 
 from __future__ import annotations
 
+import logging
+import sys
 from typing import TYPE_CHECKING
 
 import numpy as np
@@ -14,6 +16,8 @@
 import warp as wp
 from tqdm import tqdm
 
+_log = logging.getLogger(__name__)
+
 import isaaclab.sim as sim_utils
 from isaaclab import cloner
 from isaaclab.managers import EventTermCfg, ManagerTermBase, ManagerTermBaseCfg, SceneEntityCfg
@@ -368,11 +372,13 @@ def roll_once(roll_ids: torch.Tensor) -> torch.Tensor:
             # not cover all groups, and a group with no rolled envs could never fill
             all_ids = torch.arange(env.num_envs, device=env.device)
 
+            _log.info("Prefilling reset buffer: 0/%d states", num_groups * harvest_size)
             with tqdm(
                 total=num_groups * harvest_size,
                 desc="Prefilling reset buffer",
                 unit="state",
                 dynamic_ncols=True,
+                disable=not sys.stderr.isatty(),
             ) as progress:
                 while not bool((self._fill >= harvest_size).all()):
                     if max_prefill_iters is not None and iteration >= max_prefill_iters:
@@ -414,6 +420,7 @@ def roll_once(roll_ids: torch.Tensor) -> torch.Tensor:
                             self._descriptor[row : row + len(take)] = feature
                         self._fill[group] += len(take)
                         progress.update(len(take))
+            _log.info("Prefilling reset buffer: done (%d states)", num_groups * harvest_size)
             if diversity_feature is not None:
                 self._keep_most_spread(num_groups, harvest_size, buffer_size_per_group)
             if success_monitor is not None: