Dataset Viewer
Auto-converted to Parquet Duplicate
eval_id
stringlengths
31
65
puzzle_dir
stringclasses
300 values
rq
stringclasses
3 values
variant
stringclasses
7 values
image_path
stringlengths
36
62
prompt
stringlengths
1.33k
3.33k
puzzle
stringclasses
3 values
task
stringclasses
7 values
image
imagewidth (px)
799
1.28k
ground_truth_json
stringlengths
16
240
metadata_json
stringlengths
180
706
convergent_forks_6x6_2k_4d_seed17__rq1_main__n5
convergent_forks_6x6_2k_4d_seed17
rq1
main
images/convergent_forks_6x6_2k_4d_seed17/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [5, 4], "inventory": {"blue": 1, "red": 1}, "opened_doors": [], "remaining_keys": []}
{"fingerprint": "1abb0b55a985b13b", "grid_size": "6x6", "num_doors": 4, "num_keys": 2, "seed": 17, "solution_length": 20, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 5}
convergent_forks_6x6_2k_4d_seed27__rq1_main__n10
convergent_forks_6x6_2k_4d_seed27
rq1
main
images/convergent_forks_6x6_2k_4d_seed27/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 5], "inventory": {"blue": 1, "red": 1}, "opened_doors": [], "remaining_keys": []}
{"fingerprint": "c154f24160319cf7", "grid_size": "6x6", "num_doors": 4, "num_keys": 2, "seed": 27, "solution_length": 24, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 10}
convergent_forks_6x6_2k_4d_seed30__rq1_main__n15
convergent_forks_6x6_2k_4d_seed30
rq1
main
images/convergent_forks_6x6_2k_4d_seed30/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 3], "inventory": {}, "opened_doors": ["D_dual"], "remaining_keys": []}
{"fingerprint": "8bfec7f51a6e2469", "grid_size": "6x6", "num_doors": 4, "num_keys": 2, "seed": 30, "solution_length": 17, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 15}
convergent_forks_6x6_3k_4d_seed44__rq1_main__n5
convergent_forks_6x6_3k_4d_seed44
rq1
main
images/convergent_forks_6x6_3k_4d_seed44/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 0], "inventory": {"blue": 1, "red": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "9d638f850df94579", "grid_size": "6x6", "num_doors": 4, "num_keys": 3, "seed": 44, "solution_length": 25, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 5}
convergent_forks_7x7_3k_4d_seed21__rq1_main__n10
convergent_forks_7x7_3k_4d_seed21
rq1
main
images/convergent_forks_7x7_3k_4d_seed21/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 2], "inventory": {"blue": 1, "green": 1, "red": 1}, "opened_doors": [], "remaining_keys": []}
{"fingerprint": "79f18b9822134db2", "grid_size": "7x7", "num_doors": 4, "num_keys": 3, "seed": 21, "solution_length": 24, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 10}
convergent_forks_7x7_3k_4d_seed44__rq1_main__n15
convergent_forks_7x7_3k_4d_seed44
rq1
main
images/convergent_forks_7x7_3k_4d_seed44/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 3], "inventory": {"blue": 1, "green": 1, "red": 1}, "opened_doors": [], "remaining_keys": []}
{"fingerprint": "6899e9f961eac451", "grid_size": "7x7", "num_doors": 4, "num_keys": 3, "seed": 44, "solution_length": 22, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 15}
convergent_forks_7x7_3k_5d_seed10__rq1_main__n5
convergent_forks_7x7_3k_5d_seed10
rq1
main
images/convergent_forks_7x7_3k_5d_seed10/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [6, 1], "inventory": {"blue": 1, "green": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "6dbcf33037f070b4", "grid_size": "7x7", "num_doors": 5, "num_keys": 3, "seed": 10, "solution_length": 30, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 5}
convergent_forks_7x7_3k_5d_seed29__rq1_main__n10
convergent_forks_7x7_3k_5d_seed29
rq1
main
images/convergent_forks_7x7_3k_5d_seed29/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 4], "inventory": {"green": 1, "red": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "1e3875a65f97a499", "grid_size": "7x7", "num_doors": 5, "num_keys": 3, "seed": 29, "solution_length": 26, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 10}
convergent_forks_7x7_3k_5d_seed38__rq1_main__n15
convergent_forks_7x7_3k_5d_seed38
rq1
main
images/convergent_forks_7x7_3k_5d_seed38/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 4], "inventory": {"blue": 1, "green": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "dd311964fe97b709", "grid_size": "7x7", "num_doors": 5, "num_keys": 3, "seed": 38, "solution_length": 29, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 15}
convergent_forks_7x7_3k_5d_seed42__rq1_main__n5
convergent_forks_7x7_3k_5d_seed42
rq1
main
images/convergent_forks_7x7_3k_5d_seed42/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 3], "inventory": {"green": 1, "red": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "91d693d760008b23", "grid_size": "7x7", "num_doors": 5, "num_keys": 3, "seed": 42, "solution_length": 25, "sub_type": "trajectory", "template_name": "convergent_forks", "trajectory_length": 5}
decoy_key_6x6_3k_5d_seed0__rq1_main__n10
decoy_key_6x6_3k_5d_seed0
rq1
main
images/decoy_key_6x6_3k_5d_seed0/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [2, 1], "inventory": {"red": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "409652ad4cf9aa56", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 0, "solution_length": 18, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_6x6_3k_5d_seed1__rq1_main__n15
decoy_key_6x6_3k_5d_seed1
rq1
main
images/decoy_key_6x6_3k_5d_seed1/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 4], "inventory": {"blue": 1}, "opened_doors": ["D_correct_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "6451a0d06a8982c0", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 1, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_6x6_3k_5d_seed12__rq1_main__n5
decoy_key_6x6_3k_5d_seed12
rq1
main
images/decoy_key_6x6_3k_5d_seed12/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 5], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_y"], "remaining_keys": ["K_y2"]}
{"fingerprint": "773a50d766ff200b", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 12, "solution_length": 16, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_6x6_3k_5d_seed14__rq1_main__n10
decoy_key_6x6_3k_5d_seed14
rq1
main
images/decoy_key_6x6_3k_5d_seed14/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 4], "inventory": {"blue": 1}, "opened_doors": ["D_correct_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "50fbbc8bf998953c", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 14, "solution_length": 16, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_6x6_3k_5d_seed19__rq1_main__n15
decoy_key_6x6_3k_5d_seed19
rq1
main
images/decoy_key_6x6_3k_5d_seed19/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 0], "inventory": {"red": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "2d343c149fd6a22e", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 19, "solution_length": 22, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_6x6_3k_5d_seed21__rq1_main__n5
decoy_key_6x6_3k_5d_seed21
rq1
main
images/decoy_key_6x6_3k_5d_seed21/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 0], "inventory": {"blue": 1}, "opened_doors": ["D_correct_y"], "remaining_keys": ["K_y2"]}
{"fingerprint": "af22f67157d8a3ad", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 21, "solution_length": 18, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_6x6_3k_5d_seed22__rq1_main__n10
decoy_key_6x6_3k_5d_seed22
rq1
main
images/decoy_key_6x6_3k_5d_seed22/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 3], "inventory": {"red": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "edd5369a68a85a01", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 22, "solution_length": 18, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_6x6_3k_5d_seed25__rq1_main__n15
decoy_key_6x6_3k_5d_seed25
rq1
main
images/decoy_key_6x6_3k_5d_seed25/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 0], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "7699040f0de22f46", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 25, "solution_length": 18, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_6x6_3k_5d_seed26__rq1_main__n5
decoy_key_6x6_3k_5d_seed26
rq1
main
images/decoy_key_6x6_3k_5d_seed26/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 3], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "1db1dee2fa52f042", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 26, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_6x6_3k_5d_seed3__rq1_main__n10
decoy_key_6x6_3k_5d_seed3
rq1
main
images/decoy_key_6x6_3k_5d_seed3/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [2, 2], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": ["D_correct_y"], "remaining_keys": []}
{"fingerprint": "d175dce8b5a3074f", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 3, "solution_length": 18, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_6x6_3k_5d_seed39__rq1_main__n15
decoy_key_6x6_3k_5d_seed39
rq1
main
images/decoy_key_6x6_3k_5d_seed39/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [5, 1], "inventory": {}, "opened_doors": ["D_correct_y", "D_exit_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "5c5164664bbeb1d2", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 39, "solution_length": 16, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_6x6_3k_5d_seed40__rq1_main__n5
decoy_key_6x6_3k_5d_seed40
rq1
main
images/decoy_key_6x6_3k_5d_seed40/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 5], "inventory": {"yellow": 1}, "opened_doors": [], "remaining_keys": ["K_y", "K_y2"]}
{"fingerprint": "de921256bcb01844", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 40, "solution_length": 23, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_6x6_3k_5d_seed42__rq1_main__n10
decoy_key_6x6_3k_5d_seed42
rq1
main
images/decoy_key_6x6_3k_5d_seed42/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [2, 2], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": ["K_b"]}
{"fingerprint": "82f6f9c3ff0e024c", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 42, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_6x6_3k_5d_seed43__rq1_main__n15
decoy_key_6x6_3k_5d_seed43
rq1
main
images/decoy_key_6x6_3k_5d_seed43/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [5, 2], "inventory": {"blue": 1}, "opened_doors": ["D_correct_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "b30b08901ffefe3d", "grid_size": "6x6", "num_doors": 5, "num_keys": 3, "seed": 43, "solution_length": 18, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_7x7_3k_6d_seed0__rq1_main__n5
decoy_key_7x7_3k_6d_seed0
rq1
main
images/decoy_key_7x7_3k_6d_seed0/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [5, 3], "inventory": {"red": 1, "yellow": 1}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "160cc04b8078ca13", "grid_size": "7x7", "num_doors": 6, "num_keys": 3, "seed": 0, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_7x7_3k_6d_seed23__rq1_main__n10
decoy_key_7x7_3k_6d_seed23
rq1
main
images/decoy_key_7x7_3k_6d_seed23/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [6, 2], "inventory": {"blue": 1, "red": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": []}
{"fingerprint": "6180182b2fe9cde0", "grid_size": "7x7", "num_doors": 6, "num_keys": 3, "seed": 23, "solution_length": 22, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_7x7_3k_6d_seed24__rq1_main__n15
decoy_key_7x7_3k_6d_seed24
rq1
main
images/decoy_key_7x7_3k_6d_seed24/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 2], "inventory": {"blue": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "8a1b9c0430a44970", "grid_size": "7x7", "num_doors": 6, "num_keys": 3, "seed": 24, "solution_length": 22, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_7x7_3k_6d_seed27__rq1_main__n5
decoy_key_7x7_3k_6d_seed27
rq1
main
images/decoy_key_7x7_3k_6d_seed27/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 1], "inventory": {"blue": 1}, "opened_doors": [], "remaining_keys": ["K_b", "K_r"]}
{"fingerprint": "00e59206c484e38b", "grid_size": "7x7", "num_doors": 6, "num_keys": 3, "seed": 27, "solution_length": 24, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_7x7_3k_6d_seed6__rq1_main__n10
decoy_key_7x7_3k_6d_seed6
rq1
main
images/decoy_key_7x7_3k_6d_seed6/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 5], "inventory": {"blue": 1, "red": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": []}
{"fingerprint": "fa8e39095dd29f7a", "grid_size": "7x7", "num_doors": 6, "num_keys": 3, "seed": 6, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_7x7_4k_6d_seed10__rq1_main__n15
decoy_key_7x7_4k_6d_seed10
rq1
main
images/decoy_key_7x7_4k_6d_seed10/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [6, 5], "inventory": {"blue": 1, "red": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "80b61e786200a9e1", "grid_size": "7x7", "num_doors": 6, "num_keys": 4, "seed": 10, "solution_length": 23, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_7x7_4k_6d_seed30__rq1_main__n5
decoy_key_7x7_4k_6d_seed30
rq1
main
images/decoy_key_7x7_4k_6d_seed30/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [6, 1], "inventory": {"blue": 1, "red": 2}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "e390434f20a864ce", "grid_size": "7x7", "num_doors": 6, "num_keys": 4, "seed": 30, "solution_length": 21, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_7x7_4k_6d_seed38__rq1_main__n10
decoy_key_7x7_4k_6d_seed38
rq1
main
images/decoy_key_7x7_4k_6d_seed38/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 4], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": ["K_y"]}
{"fingerprint": "34f77bfb5f81759d", "grid_size": "7x7", "num_doors": 6, "num_keys": 4, "seed": 38, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_3k_7d_seed15__rq1_main__n15
decoy_key_8x8_3k_7d_seed15
rq1
main
images/decoy_key_8x8_3k_7d_seed15/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 6], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "6fb73fff964a3c97", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 15, "solution_length": 24, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_3k_7d_seed16__rq1_main__n5
decoy_key_8x8_3k_7d_seed16
rq1
main
images/decoy_key_8x8_3k_7d_seed16/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 1], "inventory": {"blue": 1}, "opened_doors": [], "remaining_keys": ["K_b", "K_y2"]}
{"fingerprint": "2de68ca900dcd6c2", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 16, "solution_length": 29, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_3k_7d_seed18__rq1_main__n10
decoy_key_8x8_3k_7d_seed18
rq1
main
images/decoy_key_8x8_3k_7d_seed18/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 6], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": ["D_correct_y"], "remaining_keys": []}
{"fingerprint": "19148b1152e556a9", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 18, "solution_length": 23, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_3k_7d_seed19__rq1_main__n15
decoy_key_8x8_3k_7d_seed19
rq1
main
images/decoy_key_8x8_3k_7d_seed19/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 0], "inventory": {"blue": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "2d6d5678fc07cb8e", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 19, "solution_length": 24, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_3k_7d_seed20__rq1_main__n5
decoy_key_8x8_3k_7d_seed20
rq1
main
images/decoy_key_8x8_3k_7d_seed20/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [2, 7], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": ["K_b"]}
{"fingerprint": "2b62fe9871a498c6", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 20, "solution_length": 23, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_3k_7d_seed23__rq1_main__n10
decoy_key_8x8_3k_7d_seed23
rq1
main
images/decoy_key_8x8_3k_7d_seed23/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [5, 2], "inventory": {"blue": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": ["K_b"]}
{"fingerprint": "56f75fdb3cee8782", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 23, "solution_length": 24, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_3k_7d_seed27__rq1_main__n15
decoy_key_8x8_3k_7d_seed27
rq1
main
images/decoy_key_8x8_3k_7d_seed27/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 4], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_r", "D_gate_y"], "remaining_keys": []}
{"fingerprint": "7b7af7a0d380badb", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 27, "solution_length": 26, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_3k_7d_seed28__rq1_main__n5
decoy_key_8x8_3k_7d_seed28
rq1
main
images/decoy_key_8x8_3k_7d_seed28/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [5, 0], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": [], "remaining_keys": ["K_y2"]}
{"fingerprint": "9fd711fbb9131348", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 28, "solution_length": 24, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_3k_7d_seed3__rq1_main__n10
decoy_key_8x8_3k_7d_seed3
rq1
main
images/decoy_key_8x8_3k_7d_seed3/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 5], "inventory": {"blue": 1}, "opened_doors": ["D_correct_y"], "remaining_keys": ["K_y2"]}
{"fingerprint": "c2889a1566668b87", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 3, "solution_length": 23, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_3k_7d_seed30__rq1_main__n15
decoy_key_8x8_3k_7d_seed30
rq1
main
images/decoy_key_8x8_3k_7d_seed30/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 4], "inventory": {"yellow": 1}, "opened_doors": ["D_correct_y", "D_gate_b"], "remaining_keys": []}
{"fingerprint": "880e4409cd554ea5", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 30, "solution_length": 20, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_3k_7d_seed43__rq1_main__n5
decoy_key_8x8_3k_7d_seed43
rq1
main
images/decoy_key_8x8_3k_7d_seed43/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 5], "inventory": {"yellow": 1}, "opened_doors": [], "remaining_keys": ["K_b", "K_r"]}
{"fingerprint": "0bc1d29de2941ac6", "grid_size": "8x8", "num_doors": 7, "num_keys": 3, "seed": 43, "solution_length": 28, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_4k_7d_seed1__rq1_main__n10
decoy_key_8x8_4k_7d_seed1
rq1
main
images/decoy_key_8x8_4k_7d_seed1/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [2, 5], "inventory": {"blue": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": ["K_b", "K_g"]}
{"fingerprint": "dcab779c4777c3a0", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 1, "solution_length": 27, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_4k_7d_seed11__rq1_main__n15
decoy_key_8x8_4k_7d_seed11
rq1
main
images/decoy_key_8x8_4k_7d_seed11/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 2], "inventory": {"green": 1}, "opened_doors": ["D_correct_r", "D_gate_b", "D_gate_g"], "remaining_keys": []}
{"fingerprint": "72aaf32f0b5e324d", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 11, "solution_length": 24, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_4k_7d_seed13__rq1_main__n5
decoy_key_8x8_4k_7d_seed13
rq1
main
images/decoy_key_8x8_4k_7d_seed13/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 7], "inventory": {"red": 1}, "opened_doors": [], "remaining_keys": ["K_b", "K_g", "K_r"]}
{"fingerprint": "1efe5c49ebabc202", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 13, "solution_length": 27, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_4k_7d_seed2__rq1_main__n10
decoy_key_8x8_4k_7d_seed2
rq1
main
images/decoy_key_8x8_4k_7d_seed2/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 1], "inventory": {"blue": 1, "red": 2}, "opened_doors": ["D_correct_r"], "remaining_keys": []}
{"fingerprint": "e4a6fec68dc3dd2b", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 2, "solution_length": 22, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_4k_7d_seed21__rq1_main__n15
decoy_key_8x8_4k_7d_seed21
rq1
main
images/decoy_key_8x8_4k_7d_seed21/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 4], "inventory": {"blue": 1}, "opened_doors": ["D_correct_r", "D_gate_b", "D_gate_g"], "remaining_keys": []}
{"fingerprint": "0f3cdd104ee7e986", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 21, "solution_length": 22, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_4k_7d_seed24__rq1_main__n5
decoy_key_8x8_4k_7d_seed24
rq1
main
images/decoy_key_8x8_4k_7d_seed24/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [0, 5], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": [], "remaining_keys": ["K_b", "K_y"]}
{"fingerprint": "58de92768313c514", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 24, "solution_length": 22, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_4k_7d_seed25__rq1_main__n10
decoy_key_8x8_4k_7d_seed25
rq1
main
images/decoy_key_8x8_4k_7d_seed25/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [7, 5], "inventory": {"blue": 1, "green": 1, "yellow": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": []}
{"fingerprint": "64e638682c75f064", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 25, "solution_length": 25, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
decoy_key_8x8_4k_7d_seed31__rq1_main__n15
decoy_key_8x8_4k_7d_seed31
rq1
main
images/decoy_key_8x8_4k_7d_seed31/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [3, 3], "inventory": {"red": 1, "yellow": 1}, "opened_doors": ["D_correct_r", "D_gate_g"], "remaining_keys": []}
{"fingerprint": "be2061909a850d79", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 31, "solution_length": 25, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 15}
decoy_key_8x8_4k_7d_seed32__rq1_main__n5
decoy_key_8x8_4k_7d_seed32
rq1
main
images/decoy_key_8x8_4k_7d_seed32/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [4, 5], "inventory": {"blue": 1, "yellow": 2}, "opened_doors": [], "remaining_keys": ["K_b"]}
{"fingerprint": "8cefc37a83115015", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 32, "solution_length": 21, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 5}
decoy_key_8x8_4k_7d_seed36__rq1_main__n10
decoy_key_8x8_4k_7d_seed36
rq1
main
images/decoy_key_8x8_4k_7d_seed36/initial.png
You are playing a grid-based puzzle called Chain Vault. ## Rules 1. You control an agent on a rectangular grid. Each cell is one of: floor, wall, key, locked door, open door, or exit. 2. The agent can move one step at a time: UP, DOWN, LEFT, or RIGHT. 3. The agent can move onto floor, keys, opened doors, and the exit...
chain_vault
rq1_main
{"agent_pos": [1, 5], "inventory": {"blue": 1, "yellow": 1}, "opened_doors": ["D_correct_r"], "remaining_keys": ["K_g"]}
{"fingerprint": "4365e23f894ae28e", "grid_size": "8x8", "num_doors": 7, "num_keys": 4, "seed": 36, "solution_length": 26, "sub_type": "trajectory", "template_name": "decoy_key", "trajectory_length": 10}
End of preview. Expand in Data Studio

RUBICON

Do Language Models Know When They've Crossed the Point of No Return?

Project website · Code and evaluation

RUBICON evaluates simulation, foresight, and causal attribution in visual puzzles with irreversible actions. This release packages the evaluation records and images for browsing and loading with Hugging Face Datasets.

Contents

Domain Mechanism Records Unique images
Chain Vault Opening doors consumes keys 700 200
Mosaic Grid Departed cells become painted and cannot be revisited 700 200
Tide Grid Actions consume oxygen as tiles change with the tide 700 200
Total 2,100 600

Each domain contains seven task variants with 100 records each. Images and puzzle instances are reused across related tasks; 2,100 records does not mean 2,100 independent puzzles. The test split names the published evaluation set; no training or validation split is supplied.

Task Question
rq1_main Simulate a given action sequence
rq2_hard_deadlock Foresight from an unsolvable state
rq2_hard_solvable Foresight from a solvable state
rq2_hard_ablation_deadlock Explicit solvability question, unsolvable state
rq2_hard_ablation_solvable Explicit solvability question, solvable state
rq3_hard Identify the first action making completion impossible
rq3_hard_ablation_final Causal attribution using the final-state image

Load the dataset

from datasets import load_dataset
import json

ds = load_dataset("ajirs/rubicon", "mosaic_grid", split="test")
row = ds[0]
image = row["image"]
prompt = row["prompt"]
ground_truth = json.loads(row["ground_truth_json"])

Select all for all 2,100 records, or chain_vault, mosaic_grid, or tide_grid for one domain. The images are embedded in the Parquet files. For reproducible experiments, pin a dataset commit using revision=.

Schema

  • eval_id: original record identifier; use (puzzle, task, eval_id) as the full key.
  • puzzle, task: domain and source task filename without its extension.
  • puzzle_dir, rq, variant: original puzzle and task metadata.
  • image: decoded puzzle image, with the original PNG bytes embedded.
  • image_path: original path relative to the domain directory.
  • prompt: unmodified evaluation prompt.
  • ground_truth_json, metadata_json: JSON strings preserving the heterogeneous original objects. Decode with json.loads.

Original files and evaluation

rubicon-original.zip contains the complete original dataset/ tree: 21 evaluation JSON files, three build-statistics JSON files, and 600 PNGs. dataset-manifest.json records SHA-256 hashes for all 624 files.

The code repository provides a checksum-verifying downloader, model runners, and scoring tools. Ground truth is included for evaluation; take care not to expose answers in model inputs or contaminate training with the evaluation set.

Limitations

Related tasks reuse puzzle instances and images. These synthetic puzzle tasks should not be treated as a comprehensive measure of real-world planning or safety. Integrity and structure checks do not constitute an independent scientific replication of all oracle labels or model results.

Downloads last month
89