token_ids listlengths 321 2.05k | text stringlengths 1.18k 9.41k | finish_reason stringclasses 2
values | targets dict | id stringlengths 37 40 | worker int64 0 0 | index int64 0 4.17k | prompt stringlengths 776 879 | prompt_ids listlengths 181 197 | teacher stringclasses 1
value | revision stringclasses 1
value | tokenizer_sha256 stringclasses 1
value | backend stringclasses 1
value | accepted bool 2
classes | verification stringclasses 1
value | sampling dict | target_semantics stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[90,198,220,328,33441,763,328,21365,264,1103,314,6572,24959,13661,359,3394,2663,504,1423,660,4752,69(...TRUNCATED) | "{\n \"problem\": \"Given a list of positive integers representing 'node values', find all unique p(...TRUNCATED) | length | {"topk_ids":[[90,71093,4754,13724,30177,5202,34521,58,8160,33441,40,760,248068,64127,68631,2164,1321(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:0 | 0 | 0 | "Create a self-contained intermediate Python exercise about backtracking over small inputs, framed a(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,1142,43075,888,2526,10804,11,44457,430,1414,(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | false | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7341
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,220,328,33441,763,328,623,279,1746,314,359,13673,15495,382,504,4078,6327,47193,19868,13,1946(...TRUNCATED) | "{\n \"problem\": \"In the game of 'Sequence Sorters', players collect numbered tiles. To win, they(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,34521,30177,68631,58,8160,248068,760,64127,623,248058,33441,1(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:1 | 0 | 1 | "Create a self-contained intermediate Python exercise about deduplication with order preservation, f(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,7438,446,1633,440,1906,44026,11,44457,430,26(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7342
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,220,328,33441,763,328,2523,513,36602,2250,13017,8806,383,264,3074,5484,13,8618,3274,682,264,(...TRUNCATED) | "{\n \"problem\": \"You are scheduling three maintenance tasks on a single machine. Each task has a(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,30177,5202,58,8160,34521,248068,760,33441,64127,68631,248058,13215(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:2 | 0 | 2 | "Create a self-contained introductory Python exercise about dynamic programming for sequences, frame(...TRUNCATED) | [248045,846,198,3886,264,638,91938,59296,12654,9856,883,8476,15019,364,22958,11,44457,430,264,2526,3(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7343
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,1,33441,763,328,2523,513,264,14023,40132,39664,264,359,37332,1737,263,663,10273,6,13849,494,(...TRUNCATED) | "{\n\"problem\": \"You are a cryptographer analyzing a 'Frequency Resonance Grid' derived from the s(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,30177,58,34521,760,8160,248068,33441,64127,68631,248058,2164,(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:3 | 0 | 3 | "Create a self-contained intermediate Python exercise about matrix traversal, framed as a numeric pu(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,5979,54409,11,44457,430,264,23311,23850,13,7(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7344
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,1,33441,763,328,2523,513,17610,264,29128,1332,3470,513,18788,1083,4574,9057,56815,82,13,1394(...TRUNCATED) | "{\n\"problem\": \"You are managing a warehouse where items are packed into perfect square pallets. (...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,58,30177,760,8160,68631,248068,34521,2164,64127,248058,40,334(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:4 | 0 | 4 | "Create a self-contained intermediate Python exercise about integer square roots, framed as an inven(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,7308,9057,19101,11,44457,430,449,14991,3377,(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7345
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,220,328,33441,763,328,21365,264,5732,4496,359,64,504,264,4047,6463,359,67,504,321,264,2100,1(...TRUNCATED) | "{\n \"problem\": \"Given a starting term 'a', a common difference 'd', and a target count 'n', gen(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,30177,34521,58,8160,68631,248068,760,64127,33441,248058,2164,(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:5 | 0 | 5 | "Create a self-contained intermediate Python exercise about arithmetic progressions, framed as a dir(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,33633,4936,880,11,44457,430,264,2050,34780,9(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7346
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,1,33441,763,328,623,279,359,65470,428,25396,18142,6,1484,4129,11,264,8240,314,7308,4216,1055(...TRUNCATED) | "{\n\"problem\": \"In the 'Chronos Coin Exchange' event stream, a sequence of integer events represe(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,30177,5202,760,34521,58,33441,8160,64127,248068,68631,2164,248058,(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:6 | 0 | 6 | "Create a self-contained introductory Python exercise about coin change, framed as an event stream. (...TRUNCATED) | [248045,846,198,3886,264,638,91938,59296,12654,9856,883,15775,2222,11,44457,430,449,1484,4129,13,717(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7347
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,1,33441,763,328,2523,513,15361,264,10031,314,7499,27437,1332,23173,4947,3875,264,3074,18677,(...TRUNCATED) | "{\n\"problem\": \"You are cleaning a dataset of transaction IDs where consecutive numbers represent(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,30177,8160,58,34521,33441,64127,248068,760,68631,40,13215,248(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:7 | 0 | 7 | "Create a self-contained intermediate Python exercise about connected components, framed as a data c(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,8173,6599,11,44457,430,264,795,15361,3274,13(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7348
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,1,33441,763,328,2523,513,17610,264,29128,1332,3470,513,28853,440,4752,27437,494,264,2574,110(...TRUNCATED) | "{\n\"problem\": \"You are managing a warehouse where items are labeled with unique IDs from a given(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,58,30177,760,8160,34521,68631,248068,64127,2523,248058,2164,4(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:8 | 0 | 8 | "Create a self-contained intermediate Python exercise about connected components, framed as an inven(...TRUNCATED) | [248045,846,198,3886,264,638,91938,27510,12654,9856,883,8173,6599,11,44457,430,449,14991,3377,13,717(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7349
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
[90,198,220,328,33441,763,328,2523,513,17610,264,29128,440,264,3074,14590,44350,1332,3470,628,1132,3(...TRUNCATED) | "{\n \"problem\": \"You are managing a warehouse with a single narrow corridor where items can only(...TRUNCATED) | stop | {"topk_ids":[[90,71093,4754,13724,5202,30177,58,34521,248068,760,64127,8160,68631,33441,2164,248058,(...TRUNCATED) | qwen35-python-math-8h-20261022-01:0:9 | 0 | 9 | "Create a self-contained advanced Python exercise about stack algorithms, framed as an inventory pro(...TRUNCATED) | [248045,846,198,3886,264,638,91938,10530,12654,9856,883,5436,24385,11,44457,430,449,14991,3377,13,71(...TRUNCATED) | Qwen/Qwen3.5-9B | c202236235762e1c871ad0ccb60c8ee5ba337b9a | a5cd9732badce41de57e6efce8302930ded1c1188c5f81feb2bd6c24c4a1941f | vllm | true | JSON shape and Python syntax only; semantics unverified | {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 7350
} | "Raw teacher distribution top-64 plus residual mass; not renormalized top-64. Native Qwen3.5 IDs, in(...TRUNCATED) |
End of preview. Expand in Data Studio
Qwen3.5 Python/math trajectories
Native pinned Qwen3.5-9B prompt/output token IDs and raw top-64 log probabilities, selected-token log probability and residual probability mass. Top-k is not renormalized. Filter accepted=true for JSON/syntax-checked examples; semantic correctness is not established. Benchmark fixtures are excluded. These token IDs are incompatible with the old shared-Qwen student vocabulary. Use runs//index.json at an immutable commit for SHA256 hashes. Generated data includes rejected outputs for audit. The old Qwen3-8B corpus remains separate.
- Downloads last month
- 591