id stringlengths 33 36 | domain stringclasses 4
values | generator stringclasses 1
value | upstream_source stringclasses 1
value | expected_answer stringlengths 0 496 | n_turns int32 1 1 | messages listlengths 3 3 |
|---|---|---|---|---|---|---|
b5b576b6-711c-5e71-a83f-9d73edc7e07a | meta_verification | DeepSeek-V4-Pro | AoPS | 1 | [
{
"role": "system",
"content": "You are an expert mathematical assistant. Provide rigorous, complete proofs.",
"reasoning_content": ""
},
{
"role": "user",
"content": "You are given a \"problem\", \"solution\", and \"solution evaluation\", and you need to assess the whether this \"solution e... | |
ultra_dsv4_cot_aops_rs1_000012148 | aops_cot | DeepSeek-V4-Pro | AoPS | \;
f^{(n)}(x)=\frac{\Gamma\!\big(\tfrac52\big)}{\Gamma\!\big(\tfrac52-n\big)}
(x+1)^{\frac32-n}
-\frac{\Gamma\!\big(\tfrac32\big)}{\Gamma\!\big(\tfrac32-n\big)}
(x+1)^{\frac12-n}\; | 1 | [
{
"role": "system",
"content": "You are an expert mathematical assistant. Provide rigorous, complete solutions.",
"reasoning_content": ""
},
{
"role": "user",
"content": "Let \\( f: [-1, 0) \\to \\mathbb{R} \\) be defined by \\( f(x) = x \\sqrt{x+1} \\). Find the \\( n \\)-th derivative \\( ... |
ultra_dsv4_cot_aops_rs1_000013026 | aops_cot | DeepSeek-V4-Pro | AoPS | \,\displaystyle
\int\frac{\tan x}{\sin^{2}x\,\sqrt{\cos^{2}x-16}}\,dx
= \frac{1}{\sqrt{15}}\,
\arctan\!\Bigg(\frac{\sqrt{\cos^{2}x-16}}{\sqrt{15}}\Bigg)
-\frac14\,
\arctan\!\Bigg(\frac{\sqrt{\cos^{2}x-16}}{4}\Bigg)+C\; | 1 | [
{
"role": "system",
"content": "You are an expert mathematical assistant. Provide rigorous, complete solutions.",
"reasoning_content": ""
},
{
"role": "user",
"content": "Evaluate the integral $\\int \\frac{\\tan(x)}{\\sin^2(x) \\sqrt{\\cos^2(x) - 16}} \\, dx$.\nPlease reason step by step, a... |
ultra_dsv4_cot_aops_rs2_000001721 | aops_cot | DeepSeek-V4-Pro | AoPS | 2 | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete so(...TRUNCATED) |
df9f7ef5-3198-5545-a6d4-a437b0e16a18 | meta_verification | DeepSeek-V4-Pro | AoPS | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete pr(...TRUNCATED) | |
007529bd-199d-5404-90ea-d3c0d9272782 | proof | DeepSeek-V4-Pro | AoPS | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete pr(...TRUNCATED) | |
909e9416-6a8d-5455-b5f9-b259afc6fa6c | verification | DeepSeek-V4-Pro | AoPS | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete pr(...TRUNCATED) | |
b1789cd2-696a-5656-aa31-416667092ed8 | proof | DeepSeek-V4-Pro | AoPS | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete pr(...TRUNCATED) | |
98a71d1f-c096-5591-8c8c-80df88167af4 | meta_verification | DeepSeek-V4-Pro | AoPS | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete pr(...TRUNCATED) | |
85b7ce21-9ecf-5c13-98ec-79d29a702238 | proof | DeepSeek-V4-Pro | AoPS | 1 | [{"role":"system","content":"You are an expert mathematical assistant. Provide rigorous, complete pr(...TRUNCATED) |
End of preview. Expand in Data Studio
yccchen-stage2-sft-v2
The stage-2 (256K long-context) SFT training mix for opd-32b (Olmo-3 32B backbone with a transplanted DeepSeek-V4 tokenizer), in the model's native DeepSeek-R1 chat format. Globally shuffled, no upsampling / no repeats — every row is a distinct solution.
- 134,698 conversations · ~6.1 B tokens (opd-32b tokenizer).
- Assistant reasoning in a separate
reasoning_contentfield (<think>…</think>), answer incontent. - Per-source system prompt; task instruction native in the user turn. No tools.
from datasets import load_dataset
ds = load_dataset("chankhavu/yccchen-stage2-sft-v2", split="train")
Mixture (all ×1, no repeats)
| source | domain | rows | system prompt |
|---|---|---|---|
| Nemotron-Math-Proofs-v2 | proof |
24,696 | "…complete proofs." |
| Nemotron-Math-Proofs-v2 | verification |
28,865 | "…complete proofs." |
| Nemotron-Math-Proofs-v2 | meta_verification |
29,176 | "…complete proofs." |
| Nemotron-SFT-Math-v4 | aops_cot |
51,961 | "…complete solutions." |
| Total | 134,698 |
- Proofs subtotal = 82,737, matching the official Nemotron-Math-Proofs-v2 train split exactly (proof 24,696 / verification 28,865 / meta_verification 29,176). No upsampling.
- The proof set already carries ~4.5 distinct solutions per problem (24,696 solutions over 5,539 problems) — natural diversity, no duplication.
- All rows globally shuffled (scatter + per-shard shuffle), sources interleaved on-disk.
- Proofs use
"You are an expert mathematical assistant. Provide rigorous, complete proofs."; AoPS CoT (numeric-answer solutions) use"…complete solutions.".
Chat format (opd-32b / DeepSeek-R1)
<｜begin▁of▁sentence｜>{system}<｜User｜>{problem}<｜Assistant｜><think>{reasoning}</think>{answer}<｜end▁of▁sentence｜>
— <think>/</think> are dedicated tokens (bos=0, eos=1). Supervised loss on the assistant span only
({reasoning}</think>{answer}<｜end▁of▁sentence｜>); everything up to <｜Assistant｜><think> is masked.
Pre-tokenized olmo-core companion: chankhavu/yccchen-stage2-olmocore-256k (docs >262144 tokens dropped).
Schema: id, domain, generator, upstream_source, expected_answer, n_turns, messages.
CC-BY-4.0 (inherits upstream Nemotron licenses).
- Downloads last month
- 283