Dataset Viewer
Auto-converted to Parquet Duplicate
index
int64
0
9k
sample_id
int64
0
7
data_source
stringclasses
1 value
prompt
stringlengths
229
2.37k
response
stringlengths
4.33k
64.6k
rewards
float64
-1
1
metadata
stringlengths
167
753
answer
stringlengths
1
21
0
0
rlve
In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). At each step, if you are currently at (x, y, z), you may move to a new (different from the current one) point of one of the following types: 1. (x', y, z) such that x AND x' = x 2. (x, y', z) such that y AND y' = y 3. (x, ...
This is a complex or challenging question, and it is difficult to provide a direct and correct answer. I need to think about it. Well, so we need to count the number of paths from (0,0,0) to (4,6,2) in 3D space with these specific movement rules, avoiding (0,6,2). Let's first make sure I understand the movement rules c...
-1
{"environment": "BannedPointSupersetPathCounting", "config": {"seed": 10000, "parameter": {"MAX_N_M_R": 6, "MAX_O": 20, "N": 4, "M": 6, "R": 2, "O": 1, "obstacles": [[0, 6, 2]], "MOD": 2879, "reference_answer": 22}, "passing_reward_threshold": 1.0}}
22
0
1
rlve
In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). At each step, if you are currently at (x, y, z), you may move to a new (different from the current one) point of one of the following types: 1. (x', y, z) such that x AND x' = x 2. (x, y', z) such that y AND y' = y 3. (x, ...
This is a complex or challenging question, and it is difficult to provide a direct and correct answer. I need to think about it. Well, so we need to count paths from (0,0,0) to (4,6,2) in 3D space with these bitwise AND move rules, avoiding (0,6,2). Let's first make sure I understand the move rules correctly. For each...
-1
{"environment": "BannedPointSupersetPathCounting", "config": {"seed": 10000, "parameter": {"MAX_N_M_R": 6, "MAX_O": 20, "N": 4, "M": 6, "R": 2, "O": 1, "obstacles": [[0, 6, 2]], "MOD": 2879, "reference_answer": 22}, "passing_reward_threshold": 1.0}}
22
0
2
rlve
In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). At each step, if you are currently at (x, y, z), you may move to a new (different from the current one) point of one of the following types: 1. (x', y, z) such that x AND x' = x 2. (x, y', z) such that y AND y' = y 3. (x, ...
This is a complex or challenging question, and it is difficult to provide a direct and correct answer. I need to think about it. Well, so we need to count the number of paths from (0,0,0) to (4,6,2) in 3D space with these specific move rules, avoiding (0,6,2). Let's first make sure I understand the move rules correctly...
-1
{"environment": "BannedPointSupersetPathCounting", "config": {"seed": 10000, "parameter": {"MAX_N_M_R": 6, "MAX_O": 20, "N": 4, "M": 6, "R": 2, "O": 1, "obstacles": [[0, 6, 2]], "MOD": 2879, "reference_answer": 22}, "passing_reward_threshold": 1.0}}
22
0
3
rlve
In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). At each step, if you are currently at (x, y, z), you may move to a new (different from the current one) point of one of the following types: 1. (x', y, z) such that x AND x' = x 2. (x, y', z) such that y AND y' = y 3. (x, ...
This is a complex or challenging question, and it is difficult to provide a direct and correct answer. I need to think about it. Well, so we need to count the number of paths from (0,0,0) to (4,6,2) in 3D space with these specific movement rules, avoiding the point (0,6,2). Let's first make sure I understand the moveme...
-1
{"environment": "BannedPointSupersetPathCounting", "config": {"seed": 10000, "parameter": {"MAX_N_M_R": 6, "MAX_O": 20, "N": 4, "M": 6, "R": 2, "O": 1, "obstacles": [[0, 6, 2]], "MOD": 2879, "reference_answer": 22}, "passing_reward_threshold": 1.0}}
22
0
4
rlve
"In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). A(...TRUNCATED)
"This is a complex or challenging question, and it is difficult to provide a direct and correct answ(...TRUNCATED)
-1
"{\"environment\": \"BannedPointSupersetPathCounting\", \"config\": {\"seed\": 10000, \"parameter\":(...TRUNCATED)
22
0
5
rlve
"In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). A(...TRUNCATED)
"This is a complex or challenging question, and it is difficult to provide a direct and correct answ(...TRUNCATED)
-1
"{\"environment\": \"BannedPointSupersetPathCounting\", \"config\": {\"seed\": 10000, \"parameter\":(...TRUNCATED)
22
0
6
rlve
"In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). A(...TRUNCATED)
"This is a complex or challenging question, and it is difficult to provide a direct and correct answ(...TRUNCATED)
-1
"{\"environment\": \"BannedPointSupersetPathCounting\", \"config\": {\"seed\": 10000, \"parameter\":(...TRUNCATED)
22
0
7
rlve
"In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (4, 6, 2). A(...TRUNCATED)
"This is a complex or challenging question, and it is difficult to provide a direct and correct answ(...TRUNCATED)
-1
"{\"environment\": \"BannedPointSupersetPathCounting\", \"config\": {\"seed\": 10000, \"parameter\":(...TRUNCATED)
22
1
0
rlve
"In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (2, 0, 6). A(...TRUNCATED)
"This is a complex or challenging question, and it is difficult to provide a direct and correct answ(...TRUNCATED)
-1
"{\"environment\": \"BannedPointSupersetPathCounting\", \"config\": {\"seed\": 10001, \"parameter\":(...TRUNCATED)
5
1
1
rlve
"In a three-dimensional space, you start at point (0, 0, 0) and want to reach the point (2, 0, 6). A(...TRUNCATED)
"This is a complex or challenging question, and it is difficult to provide a direct and correct answ(...TRUNCATED)
-1
"{\"environment\": \"BannedPointSupersetPathCounting\", \"config\": {\"seed\": 10001, \"parameter\":(...TRUNCATED)
5
End of preview. Expand in Data Studio

RLVE teacher rollouts — Qwen3-4B-Thinking-2507 (pass@8)

Teacher rollouts for on-policy distillation on the RLVE environment suite.

  • Teacher / sampler: Qwen3-4B-Thinking-2507
  • Source prompts: RLVE train split — 9000 questions across 18 environments (counting / combinatorics / optimization tasks)
  • Sampling: 8 samples/question (pass@8) = 72000 records, temperature 1.0 (sample.sh default 0.7 -> here T per run), max 16384 new tokens
  • Rewards: recomputed offline with the RLVE-Eval Gym verifier (reward in [-1, 1]: +1 correct, -0.5 wrong answer, -1 wrong/missing format). Teacher accuracy (reward>0): 23153 / 72000 = 32.2%.

Record schema (one JSON object per line)

field meaning
index question index (0-8999)
sample_id sample id within the question (0-7)
data_source "rlve"
prompt the problem statement
response the teacher's generated solution (full chain-of-thought)
rewards Gym-verifier reward in [-1, 1] (re-scored)
metadata JSON string: environment name + config (for the verifier)
answer ground-truth answer
Downloads last month
30