udpated prompts

def grounding_format_prompt(max_actions_per_step, action_sep, add_example=True):

    base_prompt = f"""You can take up to {max_actions_per_step} action(s) at a time, separated by {action_sep}.

You should first give the current state, then your thought process, and finally your answer.

The state should be in the format of {{"player":(row,column),"target":(row,column)}}

The state should be in the format of {{"player":(row1,column1),"target":(row2,column2)}}

Your response should be in the format of:

<current_state>...</current_state><think>...</think><answer>...</answer>"""

def worldmodeling_format_prompt(max_actions_per_step, action_sep, add_example=True):

    base_prompt = f"""You can take up to {max_actions_per_step} action(s) at a time, separated by {action_sep}.

You should first give your thought process, then your answer, and finally predict the next state.

The state should be in the format of {{"player":(row,column),"target":(row,column)}}

The state should be in the format of {{"player":(row1,column1),"target":(row2,column2)}}

Your response should be in the format of:

<think>...</think><answer>...</answer><next_state>...</next_state>"""

def grounding_worldmodeling_format_prompt(max_actions_per_step, action_sep, add_example=True):

    base_prompt = f"""You can take up to {max_actions_per_step} action(s) at a time, separated by {action_sep}.

You should first give the current state, then your thought process, then your answer, and finally predict the next state.

The state should be in the format of {{"player":(row,column),"target":(row,column)}}

The state should be in the format of {{"player":(row1,column1),"target":(row2,column2)}}

Your response should be in the format of:

<current_state>...</current_state><think>...</think><answer>...</answer><next_state>...</next_state>"""

def grounding_format_prompt(max_actions_per_step, action_sep, add_example=True):

    base_prompt = f"""You can take up to {max_actions_per_step} action(s) at a time, separated by {action_sep}.

You should first give the current state, then your thought process, and finally your answer.

The state should be in the format of {{"player":(row,column),"box":(row1,column1),"target":(row1,column1)}}

The state should be in the format of {{"player":(row1,column1),"box":(row2,column2),"target":(row3,column3)}}

Your response should be in the format of:

<current_state>...</current_state><think>...</think><answer>...</answer>"""

def worldmodeling_format_prompt(max_actions_per_step, action_sep, add_example=True):

    base_prompt = f"""You can take up to {max_actions_per_step} action(s) at a time, separated by {action_sep}.

You should first give your thought process, then your answer, and finally predict the next state.

The state should be in the format of {{"player":(row,column),"box":(row1,column1),"target":(row1,column1)}}

The state should be in the format of {{"player":(row1,column1),"box":(row2,column2),"target":(row3,column3)}}

Your response should be in the format of:

<think>...</think><answer>...</answer><next_state>...</next_state>"""

def grounding_worldmodeling_format_prompt(max_actions_per_step, action_sep, add_example=True):

    base_prompt = f"""You can take up to {max_actions_per_step} action(s) at a time, separated by {action_sep}.

You should first give the current state, then your thought process, then your answer, and finally predict the next state.

The state should be in the format of {{"player":(row,column),"box":(row1,column1),"target":(row1,column1)}}

The state should be in the format of {{"player":(row1,column1),"box":(row2,column2),"target":(row3,column3)}}

Your response should be in the format of:

<current_state>...</current_state><think>...</think><answer>...</answer><next_state>...</next_state>"""