tussiiiii/Qwen3-4B-AgentBench-Merged-v2-19
Text Generation • 4B • Updated • 6
Datasets:
messages listlengths 2 38 | metadata dict | mode stringclasses 3
values | task_type stringclasses 5
values | teacher_repo stringclasses 1
value | n_steps int64 0 18 | mistakes int64 0 0 | seed int64 3.41k 4.41k |
|---|---|---|---|---|---|---|---|
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3407,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,407 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 0,
"seed": 3408,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 0 | 0 | 3,408 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3409,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,409 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 4,
"seed": 3410,
"task_type": "examine_in_light",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | examine_in_light | Qwen/Qwen2.5-72B-Instruct-AWQ | 4 | 0 | 3,410 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3411,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,411 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3412,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,412 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3413,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,413 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3414,
"task_type": "cool_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | cool_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,414 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 5,
"seed": 3415,
"task_type": "examine_in_light",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | examine_in_light | Qwen/Qwen2.5-72B-Instruct-AWQ | 5 | 0 | 3,415 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "robust",
"n_steps": 7,
"seed": 3416,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | robust | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,416 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3417,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,417 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3418,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,418 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 4,
"seed": 3419,
"task_type": "examine_in_light",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | examine_in_light | Qwen/Qwen2.5-72B-Instruct-AWQ | 4 | 0 | 3,419 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3420,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,420 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3421,
"task_type": "cool_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | cool_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,421 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "explore",
"n_steps": 8,
"seed": 3422,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | explore | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,422 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3423,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,423 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "explore",
"n_steps": 11,
"seed": 3424,
"task_type": "cool_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | explore | cool_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 11 | 0 | 3,424 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "explore",
"n_steps": 5,
"seed": 3425,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | explore | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 5 | 0 | 3,425 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3426,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,426 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3427,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,427 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 4,
"seed": 3428,
"task_type": "examine_in_light",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | examine_in_light | Qwen/Qwen2.5-72B-Instruct-AWQ | 4 | 0 | 3,428 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3429,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,429 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3430,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,430 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3431,
"task_type": "cool_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | cool_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,431 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3432,
"task_type": "cool_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | cool_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,432 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3433,
"task_type": "cool_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | cool_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,433 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3434,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,434 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "robust",
"n_steps": 8,
"seed": 3435,
"task_type": "clean_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | robust | clean_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,435 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 7,
"seed": 3436,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 7 | 0 | 3,436 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3437,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,437 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3438,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,438 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3439,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,439 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 6,
"seed": 3440,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 6 | 0 | 3,440 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3441,
"task_type": "heat_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | heat_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,441 |
[
{
"content": "You are a household task-solving agent. Interact with the environment to complete the given task. At each step, you will receive an observation describing what you see. Think step by step about what to do next, then output your action.",
"role": "system"
},
{
"content": "Your task is t... | {
"mistakes": 0,
"mode": "score",
"n_steps": 8,
"seed": 3442,
"task_type": "pick_two_and_place",
"teacher_repo": "Qwen/Qwen2.5-72B-Instruct-AWQ"
} | score | pick_two_and_place | Qwen/Qwen2.5-72B-Instruct-AWQ | 8 | 0 | 3,442 |
End of preview. Expand in Data Studio
Interactive Household Task Trajectory Dataset
This dataset contains synthetic multi-step interactive task trajectories generated from a fully self-constructed environment simulator.
The data follows the same multi-turn messages structure as the
official training dataset provided by the competition organizers.
Format
messages: list of objects with:role: system | user | assistantcontent: string
- Assistant turns follow:
Think: ...Act: <action>
- Environment feedback is provided as subsequent
usermessages.
Environment & Data Generation
- All environments, tasks, and trajectories are generated from scratch.
- No external AgentBench or ALFWorld datasets were used.
- No data derived from external benchmarks was included.
Teacher LLM
- Model:
Qwen/Qwen2.5-72B-Instruct-AWQ - Runtime:
vLLM(AWQ)
Generation notes
- Actions are planned by a deterministic planner to ensure valid transitions.
- A small fraction of recovery turns ("Nothing happens.") may be injected to improve robustness.
Columns
messagestask_typeteacher_repon_stepsmistakes
- Downloads last month
- 16
Size of downloaded dataset files:
597 kB
Size of the auto-converted Parquet files:
597 kB
Number of rows:
1,000
Models trained or fine-tuned on tussiiiii/alfworld_synth_v4