CoolFace
Datasetpublic

CL-From-Nothing/rlve_teacher_topk16_20K

RLVE Teacher Top-16 Logit Data (20K) Teacher top-k logit sidecar data for continuation-style KD-SFT warmup (see compute_teacher_topk_logprobs.py / KDContinuationDataset). Each row holds, per response token, the teacher's top-16 (+ forced true token) candidate token ids and their log-probabilities, joined to the base dataset by row_id. Configs rlve_offline_20K — 20,000 rows (rlve_offline_20K_teacher_top16.parquet) rlve_rose_20K — 20,000 rows… See the full description on the dataset page: https://huggingface.co/datasets/CL-From-Nothing/rlve_teacher_topk16_20K.

sourceHugging Facemitupdated 2mo agoView on Hugging Face
0likes26downloads
Dataset Card

RLVE Teacher Top-16 Logit Data (20K)

Teacher top-k logit sidecar data for continuation-style KD-SFT warmup (see compute_teacher_topk_logprobs.py / KDContinuationDataset).

Each row holds, per response token, the teacher's top-16 (+ forced true token) candidate token ids and their log-probabilities, joined to the base dataset by row_id.

Configs

  • —rlve_offline_20K — 20,000 rows (rlve_offline_20K_teacher_top16.parquet)
  • —rlve_rose_20K — 20,000 rows (rlve_rose_20K_teacher_top16.parquet)

Schema

columntypedescription
row_idint64join key to the base RLVE dataset
response_token_lenint32number of response tokens
teacher_topk_idslist[list[int64]](seq, top_k+1) top-k candidate token ids (true token forced in)
teacher_topk_logprobslist[list[float32]](seq, top_k+1) teacher log-probs for those candidates

top_k = 16, width = 17.