diff --git a/examples/rl/grpo/gsm8k/run_gemma2_2b.sh b/examples/rl/grpo/gsm8k/run_gemma2_2b.sh
index ec58572b0..db2b2b503 100755
--- a/examples/rl/grpo/gsm8k/run_gemma2_2b.sh
+++ b/examples/rl/grpo/gsm8k/run_gemma2_2b.sh
@@ -34,7 +34,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
@@ -67,4 +67,3 @@ python3 -m tunix.cli.grpo_main \
   rollout_config.total_generation_steps=$total_generation_steps \
   rollout_config.max_prompt_length=$max_prompt_length \
   "$@"
-
diff --git a/examples/rl/grpo/gsm8k/run_gemma3_12b.sh b/examples/rl/grpo/gsm8k/run_gemma3_12b.sh
index cda13adb1..6f21f490a 100755
--- a/examples/rl/grpo/gsm8k/run_gemma3_12b.sh
+++ b/examples/rl/grpo/gsm8k/run_gemma3_12b.sh
@@ -28,7 +28,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
diff --git a/examples/rl/grpo/gsm8k/run_gemma3_1b.sh b/examples/rl/grpo/gsm8k/run_gemma3_1b.sh
index 54f0bc2e0..50633eb06 100755
--- a/examples/rl/grpo/gsm8k/run_gemma3_1b.sh
+++ b/examples/rl/grpo/gsm8k/run_gemma3_1b.sh
@@ -28,7 +28,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
@@ -55,4 +55,3 @@ python3 -m tunix.cli.grpo_main \
   rl_training_config.max_steps=$max_steps \
   rl_training_config.metrics_logging_options.log_dir="/tmp/tensorboard/grpo_gemma3_1b" \
   "$@"
-
diff --git a/examples/rl/grpo/gsm8k/run_gemma3_4b.sh b/examples/rl/grpo/gsm8k/run_gemma3_4b.sh
index 5ffb8e1cd..c5b4d6be2 100755
--- a/examples/rl/grpo/gsm8k/run_gemma3_4b.sh
+++ b/examples/rl/grpo/gsm8k/run_gemma3_4b.sh
@@ -28,7 +28,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
@@ -55,4 +55,3 @@ python3 -m tunix.cli.grpo_main \
   rl_training_config.max_steps=$max_steps \
   rl_training_config.metrics_logging_options.log_dir="/tmp/tensorboard/grpo_gemma3_4b" \
   "$@"
-
diff --git a/examples/rl/grpo/gsm8k/run_llama3.2_1b.sh b/examples/rl/grpo/gsm8k/run_llama3.2_1b.sh
index 236b2a3dd..fcc28a84b 100755
--- a/examples/rl/grpo/gsm8k/run_llama3.2_1b.sh
+++ b/examples/rl/grpo/gsm8k/run_llama3.2_1b.sh
@@ -28,7 +28,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
diff --git a/examples/rl/grpo/gsm8k/run_qwen3_simplereward.sh b/examples/rl/grpo/gsm8k/run_qwen3_simplereward.sh
index 8fc4e55a0..19b4129e8 100644
--- a/examples/rl/grpo/gsm8k/run_qwen3_simplereward.sh
+++ b/examples/rl/grpo/gsm8k/run_qwen3_simplereward.sh
@@ -30,7 +30,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 max_steps=$(printf "%.0f" "$max_steps_float")
 warmup_steps=$(awk "BEGIN {printf \"%.0f\", $warmup_ratio * $max_steps}")
 
@@ -99,4 +99,3 @@ python3 -m tunix.cli.grpo_main \
   grpo_config.epsilon=0.2 \
   reward_functions="['tunix/cli/reward_fn/simple_math.py']" \
   "$@"
-
diff --git a/examples/rl/grpo/gsm8k/verl_compatible/run_llama3.2_1b.sh b/examples/rl/grpo/gsm8k/verl_compatible/run_llama3.2_1b.sh
index 4f4d5e1ee..cf81cb024 100755
--- a/examples/rl/grpo/gsm8k/verl_compatible/run_llama3.2_1b.sh
+++ b/examples/rl/grpo/gsm8k/verl_compatible/run_llama3.2_1b.sh
@@ -28,7 +28,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
diff --git a/examples/sft/mtnt/run_gemma3_4b.sh b/examples/sft/mtnt/run_gemma3_4b.sh
index 5ffb8e1cd..c5b4d6be2 100755
--- a/examples/sft/mtnt/run_gemma3_4b.sh
+++ b/examples/sft/mtnt/run_gemma3_4b.sh
@@ -28,7 +28,7 @@ echo "  Num Epochs: $num_train_epochs"
 echo "  Warmup Ratio: $warmup_ratio"
 echo "  Train Fraction: $train_fraction"
 
-max_steps_float=$(awk "BEGIN {print $batch_size * $num_batches * $num_train_epochs * $train_fraction}")
+max_steps_float=$(awk "BEGIN {print $num_batches * $num_train_epochs * $train_fraction}")
 
 max_steps=$(printf "%.0f" "$max_steps_float")
 
@@ -55,4 +55,3 @@ python3 -m tunix.cli.grpo_main \
   rl_training_config.max_steps=$max_steps \
   rl_training_config.metrics_logging_options.log_dir="/tmp/tensorboard/grpo_gemma3_4b" \
   "$@"
-