Datasets:
Tasks:
Text Classification
Formats:
parquet
Languages:
English
Size:
10K - 100K
Tags:
process-reward-model
value-model
mathematical-reasoning
theorem-proving
monte-carlo
tree-search
License:
messages listlengths 2 2 | enable_thinking bool 1
class | id stringclasses 889
values | source stringclasses 2
values | sample_index int64 0 3 | step_idx int64 1 9 | num_steps int64 8 8 | label stringclasses 1
value | correct bool 2
classes | loss_weight float64 1 1 | reward float64 0 1 |
|---|---|---|---|---|---|---|---|---|---|---|
[
{
"role": "user",
"content": "# Role\nYou are forecasting the normalized rubric value of a fixed mathematical response\nstate. The state was produced by a search policy and is either incomplete or a\ncompleted response.\n\nFor an incomplete state, the rolling working memory and current reasoning chunk are\n... | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 1 | 8 | yes | false | 1 | 0 |
[
{
"role": "user",
"content": "# Role\nYou are forecasting the normalized rubric value of a fixed mathematical response\nstate. The state was produced by a search policy and is either incomplete or a\ncompleted response.\n\nFor an incomplete state, the rolling working memory and current reasoning chunk are\n... | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 2 | 8 | yes | false | 1 | 0.017857 |
[
{
"role": "user",
"content": "# Role\nYou are forecasting the normalized rubric value of a fixed mathematical response\nstate. The state was produced by a search policy and is either incomplete or a\ncompleted response.\n\nFor an incomplete state, the rolling working memory and current reasoning chunk are\n... | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 3 | 8 | yes | false | 1 | 0 |
[
{
"role": "user",
"content": "# Role\nYou are forecasting the normalized rubric value of a fixed mathematical response\nstate. The state was produced by a search policy and is either incomplete or a\ncompleted response.\n\nFor an incomplete state, the rolling working memory and current reasoning chunk are\n... | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 4 | 8 | yes | false | 1 | 0.017857 |
[
{
"role": "user",
"content": "# Role\nYou are forecasting the normalized rubric value of a fixed mathematical response\nstate. The state was produced by a search policy and is either incomplete or a\ncompleted response.\n\nFor an incomplete state, the rolling working memory and current reasoning chunk are\n... | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 5 | 8 | yes | false | 1 | 0.017857 |
[
{
"role": "user",
"content": "# Role\nYou are forecasting the normalized rubric value of a fixed mathematical response\nstate. The state was produced by a search policy and is either incomplete or a\ncompleted response.\n\nFor an incomplete state, the rolling working memory and current reasoning chunk are\n... | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 6 | 8 | yes | false | 1 | 0.071429 |
[{"role":"user","content":"# Role\nYou are forecasting the normalized rubric value of a fixed mathem(...TRUNCATED) | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 7 | 8 | yes | false | 1 | 0 |
[{"role":"user","content":"# Role\nYou are forecasting the normalized rubric value of a fixed mathem(...TRUNCATED) | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 0 | 8 | 8 | yes | false | 1 | 0.017857 |
[{"role":"user","content":"# Role\nYou are forecasting the normalized rubric value of a fixed mathem(...TRUNCATED) | false | 002dea139b3d9b09 | tcs_direction_mc8_endpoint | 0 | 9 | 8 | yes | false | 1 | 0 |
[{"role":"user","content":"# Role\nYou are forecasting the normalized rubric value of a fixed mathem(...TRUNCATED) | false | 002dea139b3d9b09 | tcs_direction_mc8_partial | 1 | 1 | 8 | yes | false | 1 | 0 |
End of preview. Expand in Data Studio
TCS Direction MC Value: Solution Label
Partial rows include 3 same-problem endpoint solution references; reference reward labels are included. Endpoint rows use the explicit no-context sentinel.
Each row is a two-message conversation ending in the literal assistant target yes. Training uses the dense reward column as the soft target P(correct); correct is only the legacy boolean projection. Hidden model reasoning is not included in endpoint state.
Identity
- Source: asingh15/tcs-qwen36-27b-direction-rollouts-partial-1017-mc8
- Source revision:
685b614552d9ba5e2f27d6aeee1ff9d6e83c9635 - Run ID:
tcs_qwen36_27b_direction_beam_full1801_mc8_20260815 - Arm:
solution_label - Verification file SHA-256:
7ddab749f98b251aea5d3ebcdb4137fd579a6858854f756f60ea4043dd032e38 - Verification contract SHA-256:
329c0fc6e175e3baf139214bee239fe956ef4158a598ee50d21bbcdb21c77279 - Maximum observed tokens:
29650 - Validated token cap:
49152
Splits
| Split | Rows | Bytes | SHA-256 |
|---|---|---|---|
| train | 32,004 | 65,906,540 | 8f3d221bdebbbb861990705b6afaf855912a087d7e39ebae74ba34f59a6660a0 |
| validation | 4,608 | 9,403,527 | ad28e83723245a3ead1f4cb87aeb4a86efec4c689f39effaea8324b170f690ee |
dataset_provenance.json, verification.json, and _SUCCESS.json pin the builder contract, source artifacts, cross-arm row alignment, and published file identities.
- Downloads last month
- 66