File size: 1,316 Bytes
c1fb314 0c3e81d c1fb314 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 | ---
license: mit
---
# DeepScaleR-Easy-10k
RL training set of 10,000 DeepScaleR questions that Gemma-3-4B-IT solved **4/4**
(temp 1.0, 4 samples, LENIENT math_verify grading), split train 9,500 / val 500.
Creation (seed 42 throughout):
1. Gemma-3-4B-IT generated 4 responses per DeepScaleR question (40,315 rows).
2. **Dedup v2**: questions grouped under max-strength normalization (case,
whitespace, prose punctuation, LaTeX formatting); duplicate groups whose gold
answers are not all math_verify-equivalent were removed entirely (167 groups /
567 rows — label conflicts); one random copy kept per surviving group
-> 38,796 unique questions. Script: `rl-distill-scripts/data/dedup_deepscaler_it_gen.py`.
3. 10,000 sampled uniformly from the lenient 4/4 bucket (10,053 available).
4. 500 sampled uniformly as validation; remaining 9,500 are train.
Files: `*_train.parquet` and `*_val500.parquet` (500 unique held-out questions).
For mean@16 validation, replicate the val rows 16x locally (verl samples one
response per row). verl format: data_source "math" (routes to the repo's
math_verify scorer), prompt = single user message, reward_model.ground_truth,
extra_info carries lenient/strict pass counts.
Built by `rl-distill-scripts/data/build_deepscaler_easy_medium.py` in
JasonWei05/rl-distill.
|