diff --git a/evaluation/benchmarks/logic_reasoning/run_infer.py b/evaluation/benchmarks/logic_reasoning/run_infer.py
index 87334de0e4..0a1447f061 100644
--- a/evaluation/benchmarks/logic_reasoning/run_infer.py
+++ b/evaluation/benchmarks/logic_reasoning/run_infer.py
@@ -272,7 +272,7 @@ if __name__ == '__main__':
         default='ProofWriter',
     )
     parser.add_argument(
-        '--data_split',
+        '--data-split',
         type=str,
         help='data split to evaluate on {validation}',  # right now we only support validation split
         default='validation',
diff --git a/evaluation/benchmarks/scienceagentbench/run_infer.py b/evaluation/benchmarks/scienceagentbench/run_infer.py
index 7e7c7919c0..db4abf0f48 100644
--- a/evaluation/benchmarks/scienceagentbench/run_infer.py
+++ b/evaluation/benchmarks/scienceagentbench/run_infer.py
@@ -251,7 +251,7 @@ If the program uses some packages that are incompatible, please figure out alter
 if __name__ == '__main__':
     parser = get_parser()
     parser.add_argument(
-        '--use_knowledge',
+        '--use-knowledge',
         type=str,
         default='false',
         choices=['true', 'false'],
diff --git a/evaluation/benchmarks/scienceagentbench/scripts/run_infer.sh b/evaluation/benchmarks/scienceagentbench/scripts/run_infer.sh
index e8abf58e03..945f7dccda 100755
--- a/evaluation/benchmarks/scienceagentbench/scripts/run_infer.sh
+++ b/evaluation/benchmarks/scienceagentbench/scripts/run_infer.sh
@@ -35,7 +35,7 @@ echo "MODEL_CONFIG: $MODEL_CONFIG"
 COMMAND="poetry run python evaluation/benchmarks/scienceagentbench/run_infer.py \
   --agent-cls $AGENT \
   --llm-config $MODEL_CONFIG \
-  --use_knowledge $USE_KNOWLEDGE \
+  --use-knowledge $USE_KNOWLEDGE \
   --max-iterations 30 \
   --eval-num-workers $NUM_WORKERS \
   --eval-note $OPENHANDS_VERSION" \
diff --git a/evaluation/benchmarks/toolqa/README.md b/evaluation/benchmarks/toolqa/README.md
index b6b25da43b..b72faca947 100644
--- a/evaluation/benchmarks/toolqa/README.md
+++ b/evaluation/benchmarks/toolqa/README.md
@@ -11,7 +11,7 @@ Please follow instruction [here](../../README.md#setup) to setup your local deve
 Make sure your Docker daemon is running, then run this bash script:
 
 ```bash
-bash evaluation/benchmarks/toolqa/scripts/run_infer.sh [model_config] [git-version] [agent] [eval_limit] [dataset] [hardness] [wolfram_alpha_appid]
+bash evaluation/benchmarks/toolqa/scripts/run_infer.sh [model_config] [git-version] [agent] [eval_limit] [dataset] [hardness] [wolfram-alpha-appid]
 ```
 
 where `model_config` is mandatory, while all other arguments are optional.
@@ -32,7 +32,7 @@ By default, the script evaluates 1 instance.
 
 `hardness`, the hardness to evaluate. You could choose from `easy` and `hard`. The default is `easy`.
 
-`wolfram_alpha_appid` is an optional argument. When given `wolfram_alpha_appid`, the agent will be able to access Wolfram Alpha's APIs.
+`wolfram-alpha-appid` is an optional argument. When given `wolfram-alpha-appid`, the agent will be able to access Wolfram Alpha's APIs.
 
 Note: in order to use `eval_limit`, you must also set `agent`; in order to use `dataset`, you must also set `eval_limit`; in order to use `hardness`, you must also set `dataset`.
 
diff --git a/evaluation/benchmarks/toolqa/run_infer.py b/evaluation/benchmarks/toolqa/run_infer.py
index 32a830e2a6..f88163a048 100644
--- a/evaluation/benchmarks/toolqa/run_infer.py
+++ b/evaluation/benchmarks/toolqa/run_infer.py
@@ -171,7 +171,7 @@ if __name__ == '__main__':
         default='easy',
     )
     parser.add_argument(
-        '--wolfram_alpha_appid',
+        '--wolfram-alpha-appid',
         type=str,
         help='wolfram alpha appid to use for wolfram alpha related tests',
         default='YOUR_WOLFRAMALPHA_APPID',
diff --git a/evaluation/benchmarks/toolqa/scripts/run_infer.sh b/evaluation/benchmarks/toolqa/scripts/run_infer.sh
index 4760613431..7f5635dd9f 100755
--- a/evaluation/benchmarks/toolqa/scripts/run_infer.sh
+++ b/evaluation/benchmarks/toolqa/scripts/run_infer.sh
@@ -53,7 +53,7 @@ COMMAND="poetry run python evaluation/benchmarks/toolqa/run_infer.py \
   --max-iterations 30 \
   --dataset $DATASET \
   --hardness $HARDNESS \
-  --wolfram_alpha_appid $WOLFRAM_APPID\
+  --wolfram-alpha-appid $WOLFRAM_APPID\
   --data-split validation \
   --eval-num-workers $NUM_WORKERS \
   --eval-note ${OPENHANDS_VERSION}_${LEVELS}"