state listlengths 2 24 | action dict | next_obs listlengths 2 2 | state_len int64 1.88k 3.25k | temperature float64 1 1 | reward float64 0.13 1 | done bool 2
classes | truncated bool 1
class | timestep int64 0 11 | try_step int64 0 0 | batch_id int64 99 99 | unique_data_sample_id int64 0 79 | generation_id int64 0 0 | split stringclasses 1
value | return_ float64 0.13 1 | advantage float64 0.13 1 | return_is_computed bool 1
class | advantage_is_computed bool 1
class | constant_reward_group bool 1
class | tools listlengths 11 11 | action_prob float64 0.13 1 | system_prompt null |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I need to find the broom in the bedroom. First, I'll take a look around to see if the broom is visible or if it's inside any container or on a supporter. This will help me locate it properly.\n\n<tool_call>\n{\"name\": \"look\", \"arguments\": {}}\n</tool_call>",
"role": "assistant"
} | [
{
"content": "-= Bedroom =-\nYou are in a bedroom. A typical kind of place.\n\n\n\nThere is an unblocked exit to the east. There is an unguarded exit to the west.\n\n\n> -= Bedroom =-0/... | 1,900 | 1 | 1 | false | false | 0 | 0 | 99 | 0 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "The broom might be in a different room, so I'll explore the west exit to find it. Moving west could lead to a room where the broom is located or stored. I'll check that area next.\n\n<tool_call>\n{\"name\": \"go\", \"arguments\": {\"direction\": \"west\"}}\n</tool_call>",
"role": "assistant"
} | [
{
"content": "-= Basement =-\nYou arrive in a basement. A normal one.\n\n\n\nYou don't like doors? Why not try going east, that entranceway is unguarded. You don't like doors? Why not try going south, that entranceway is unguarded.\n\nThere is a broom and a fly larva on the floor.\n\n\n> ... | 2,134 | 1 | 1 | false | false | 1 | 0 | 99 | 0 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I found the broom on the floor of the basement. I'll take it to complete the task. The objective is to \"take broom,\" and since I now have it in my inventory, the task should be considered complete.\n\n<tool_call>\n{\"name\": \"take\", \"arguments\": {\"item\": \"broom\"}}\n</tool_call>",
"role": "as... | [
{
"content": "",
"role": "assistant"
},
{
"content": "",
"role": "assistant"
}
] | 2,273 | 1 | 1 | true | false | 2 | 0 | 99 | 0 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I need to explore the bar environment to find the \"latchkey\" as per the new objective. First, I'll use the `look` command to examine the current room and see if the latchkey is visible or if it's in a container or on a surface. This will help me locate the item.\n\n<tool_call>\n{\"name\": \"look\", \"... | [
{
"content": "-= Bar =-\nYou've entered a bar.\n\n\n\nYou don't like doors? Why not try going north, that entranceway is unguarded.\n\n\n> -= Bar =-0/2\n\n## Current State:\n\n{\n \"cu... | 1,886 | 1 | 1 | false | false | 0 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "The bar has a north entrance that is unguarded. I'll move north to explore further, as the latchkey might be in a new area. This is a logical step to continue my search for the latchkey.\n\n<tool_call>\n{\"name\": \"go\", \"arguments\": {\"direction\": \"north\"}}\n</tool_call>",
"role": "assistant"
} | [
{
"content": "-= Kitchen =-\nWell, here we are in the kitchen.\n\n\n\nThere is an exit to the east. Don't worry, it is unblocked. There is an unblocked exit to the south.\n\n\n> -= Kitc... | 2,123 | 1 | 1 | false | false | 1 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I'm now in the kitchen. I need to check the surroundings more carefully to locate the \"latchkey.\" Using `look` will reveal any visible items or objects, including whether the latchkey is on the floor, in a container, or near a door. This step is essential to progress toward completing the task.\n\n<to... | [
{
"content": "-= Kitchen =-\nWell, here we are in the kitchen.\n\n\n\nThere is an exit to the east. Don't worry, it is unblocked. There is an unblocked exit to the south.\n\n\n> -= Kitc... | 2,222 | 1 | 1 | false | false | 2 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "The kitchen has an exit to the east. I'll explore that direction to search for the latchkey. It might be in a different room, such as a pantry or storage area, which could contain the latchkey. Moving east is a logical step to continue my search.\n\n<tool_call>\n{\"name\": \"go\", \"arguments\": {\"dire... | [
{
"content": "-= Bedchamber =-\nYou arrive in a bedchamber. An ordinary kind of place. The room seems oddly familiar, as though it were only superficially different from the other rooms in the building.\n\n\n\nYou don't like doors? Why not try going south, that entranceway is unguarded. There is an unblocked ex... | 2,314 | 1 | 1 | false | false | 3 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I'm now in the bedchamber. I need to check for any visible items, especially the \"latchkey,\" as it might be in this room or nearby. Using `look` will help me determine if the latchkey is visible or if I need to explore further. This is crucial for completing the task.\n\n<tool_call>\n{\"name\": \"look... | [
{
"content": "-= Bedchamber =-\nYou arrive in a bedchamber. An ordinary kind of place. The room seems oddly familiar, as though it were only superficially different from the other rooms in the building.\n\n\n\nYou don't like doors? Why not try going south, that entranceway is unguarded. There is an unblocked ex... | 2,466 | 1 | 1 | false | false | 4 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "The bedchamber has a south exit that is unguarded. I'll move south to continue exploring, as the latchkey might be in a new area. This step will help me cover more ground and increase my chances of finding the latchkey. I'll keep exploring systematically until I locate it.\n\n<tool_call>\n{\"name\": \"g... | [
{
"content": "-= Workshop =-\nYou've just sauntered into a workshop.\n\n\n\nYou don't like doors? Why not try going east, that entranceway is unblocked. There is an unguarded exit to the north.\n\nThere is a gummy bear and a latchkey on the floor.\n\n\n> ... | 2,558 | 1 | 1 | false | false | 5 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I have found the latchkey on the floor of the workshop. The task requires me to take the latchkey. After taking it, I will verify the score to confirm if the task is complete. The score should update to 1 when the objective is fulfilled.\n\n<tool_call>\n{\"name\": \"take\", \"arguments\": {\"item\": \"l... | [
{
"content": "",
"role": "assistant"
},
{
"content": "",
"role": "assistant"
}
] | 2,623 | 1 | 1 | true | false | 6 | 0 | 99 | 1 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I need to explore the shower room to find any clues or items related to the teapot or refrigerator. The passkey might be relevant here, or there could be a path to a storage area. I'll take a full look around to gather information.\n\n<tool_call>\n{\"name\": \"look\", \"arguments\": {}}\n</tool_call>",
... | [
{
"content": "-= Shower =-\nYou find yourself in a shower. A normal kind of place. You begin to take stock of what's in the room.\n\n\n\nYou need an unblocked exit? You should try going east. You need an unblocked exit? You should try going south.\n\n\n> ... | 1,938 | 1 | 1 | false | false | 0 | 0 | 99 | 2 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "The shower has exits to the east and south. I need to explore further to find the refrigerator or any related items. Going east might lead to a new area with the teapot. I'll proceed to the east first.\n\n<tool_call>\n{\"name\": \"go\", \"arguments\": {\"direction\": \"east\"}}\n</tool_call>",
"role":... | [
{
"content": "-= Cellar =-\nYou arrive in a cellar. An usual kind of place.\n\n\n\nThere is an exit to the south. Don't worry, it is unguarded. You need an unblocked exit? You should try going west.\n\n\n> ... | 2,214 | 1 | 1 | false | false | 1 | 0 | 99 | 2 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I'm in the cellar again. There's an exit to the south and a suggestion to go west. I need to find the refrigerator or the teapot. Going west might lead to a storage area or the refrigerator. I'll try going west now to explore further.\n\n<tool_call>\n{\"name\": \"go\", \"arguments\": {\"direction\": \"s... | [
{
"content": "-= Chamber =-\nYou are in a chamber. A standard kind of place.\n\n\n\nYou don't like doors? Why not try going north, that entranceway is unguarded.\n\nThere is a passkey on the floor.\n\n\n> ... | 2,277 | 1 | 0.357801 | false | false | 2 | 0 | 99 | 2 | 0 | train_all | 0.357801 | 0.357801 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 0.357801 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I found another passkey in the chamber. This suggests there might be multiple passkeys or a pattern in the environment. However, since the goal is to recover the teapot from the refrigerator, I need to find the refrigerator itself. I should continue exploring to locate it. I'll take this passkey to keep... | [
{
"content": "",
"role": "assistant"
},
{
"content": "",
"role": "assistant"
}
] | 2,358 | 1 | 1 | true | false | 3 | 0 | 99 | 2 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I need to explore the studio more thoroughly to find the refrigerator and the teapot. The environment has exits to north, south, and west, but I don't know what's inside yet. I'll look around to identify any visible objects or clues that might help me find the refrigerator or the teapot.\n\n<tool_call>\... | [
{
"content": "-= Studio =-\nThis might come as a shock to you, but you've just entered a studio.\n\n\n\nYou don't like doors? Why not try going north, that entranceway is unblocked. You need an unguarded exit? You should try going south. You don't like doors? Why not try going west, that entranceway is unblocke... | 1,972 | 1 | 0.931063 | false | false | 0 | 0 | 99 | 3 | 0 | train_all | 0.931063 | 0.931063 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 0.931063 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I'll try going north first, as the entranceway is unblocked. This might lead to a room with the refrigerator or the teapot. I need to explore systematically to locate the objective.\n\n<tool_call>\n{\"name\": \"go\", \"arguments\": {\"direction\": \"north\"}}\n</tool_call>",
"role": "assistant"
} | [
{
"content": "-= Cookery =-\nYou arrive in a cookery. A normal kind of place. You can barely contain your excitement.\n\n\n\nYou don't like doors? Why not try going south, that entranceway is unguarded.\n\nThere is a passkey on the floor.\n\n\n> ... | 2,293 | 1 | 1 | false | false | 1 | 0 | 99 | 3 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I found another passkey in the cookery. I now have two passkeys, but I still need to find the refrigerator and recover the teapot. I'll keep the passkey for future use, especially if I need to unlock a door or container. I'll now explore further to locate the refrigerator or teapot.\n\n<tool_call>\n{\"n... | [
{
"content": "",
"role": "assistant"
},
{
"content": "",
"role": "assistant"
}
] | 2,330 | 1 | 0.980457 | true | false | 2 | 0 | 99 | 3 | 0 | train_all | 0.980457 | 0.980457 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 0.980457 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I am now in the Chamber. I need to explore the environment thoroughly to locate the fondue. First, I'll use the \"look\" command to get a complete description of the room and any visible items or objects. This will help me identify where the fondue might be located or if there are any clues to guide me.... | [
{
"content": "-= Chamber =-\nYou're now in the chamber. You decide to just list off a complete list of everything you see in the room, because hey, why not?\n\n\n\nYou need an unguarded exit? You should try going east. You don't like doors? Why not try going north, that entranceway is unguarded.\n\n\n> ... | 1,960 | 1 | 1 | false | false | 0 | 0 | 99 | 4 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "The chamber has an unguarded exit to the east. I will move east to explore the new area, as this might lead to the fondue. Before moving, I should confirm that the east direction is accessible and contains relevant items or locations. Proceeding east is the logical next step to find the fondue.\n\n<tool... | [
{
"content": "-= Cookhouse =-\nYou have fallen into the most typical of all possible cookhouses.\n\n\n\nYou need an unguarded exit? You should try going west.\n\nThere is a fondue on the floor.\n\n\n> ... | 2,274 | 1 | 1 | false | false | 1 | 0 | 99 | 4 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
[
{
"content": "You are a helpful tool-calling assistant.\n\nYou are an agent playing TextWorld, a text-based adventure game where you are in a randomly generated maze and must find a specific object.\nYou need to explore different rooms to find the target object.\n\n## Rules:\n- You need to unlock a locked door ... | {
"content": "I found the fondue on the floor in the cookhouse. The task requires me to take the fondue. I will pick it up and complete the objective. Once the fondue is in my inventory, the task is fulfilled, and the score should update to 1.\n\n<tool_call>\n{\"name\": \"take\", \"arguments\": {\"item\": \"fondue\"}... | [
{
"content": "",
"role": "assistant"
},
{
"content": "",
"role": "assistant"
}
] | 2,321 | 1 | 1 | true | false | 2 | 0 | 99 | 4 | 0 | train_all | 1 | 1 | true | true | false | [
{
"description": "Describe the current room.",
"name": "look",
"parameters": {
"properties": {
"container": null,
"direction": null,
"item": null,
"key": null,
"source": null,
"supporter": null,
"target": null
},
"required": [],
... | 1 | null |
End of preview. Expand in Data Studio
Act-PRM Rollout Dataset
Run Metadata
- run_name:
act-prm-v2-ia=0-rr=0-ec=act_prm_tw_treasure_easy_gpt5m-gc=aprm_qwen3_ap-tc=aprm_for_sft100-rbc=default-mc=hf_qwen3_4b_inst_2507-lc=r8_a16_qkvo-nf=0-ao=1-ho=1-gs=8-bs=8-lr=4e_05-ns=4-s=0-r=gpt5m_hd_0 - run_cmd:
main_pytorch.py --env_config act_prm/tw_treasure_easy_gpt5m --model_config hf_qwen3_4b_inst_2507 --lora_config r8_a16_qkvo --generator_config aprm_qwen3_ap --trainer_config aprm_for_sft100 --replay_buffer_config default --log_path ./logs --actions_only --hide_observations --batch_size 8 --group_size 8 --learning_rate 4e-5 --num_substeps 4 --seed 0 --replicate gpt5m_hd_0 --verbose - batch_idx:
99 - split:
train_all - trajectory_key:
think_act_policy - num_trajectories:
348 - num_episodes:
348 - group_size:
8 - batch_size:
8
- Downloads last month
- 27