---
title: Logics-STEM-SFT-Dataset Overview
url: https://www.emergentmind.com/topics/logics-stem-sft-dataset
type: topic
---

# Logics-STEM-SFT-Dataset Overview

Logics-STEM-SFT-Dataset is a large-scale supervised fine-tuning corpus released with the reasoning model Logics-STEM. It consists of question–chain-of-thought–answer triples for STEM reasoning, with a scale of approximately 10 million instances in the full release and 2.2 million instances in a stratified downsample. Each record is text-only, with images excluded, and is annotated with metadata such as domain, educational level, and answer type. The dataset was constructed through a five-stage curation pipeline—annotation, deduplication, decontamination, distillation, and stratified sampling—and is used within a failure-driven post-training framework that combines targeted retrieval and synthetic data generation to improve reasoning performance on external STEM benchmarks [2601.01562].

## 1. Dataset identity and scope

Logics-STEM-SFT-Dataset is a **training corpus**, not a benchmark with internal evaluation splits. The release includes two versions: **Logics-STEM-SFT-10M**, containing approximately 10 million question–CoT–answer triples, and **Logics-STEM-SFT-2.2M**, a stratified downsample retaining 2.2 million instances. The stated purpose is post-training for reasoning models in STEM domains, especially for supervised fine-tuning and subsequent second-stage SFT or reinforcement learning [2601.01562].

The scope is broader than mathematics alone. The metadata explicitly covers **Math, Physics, Chemistry, Biology/Medicine, Computer Science, and Engineering**, and the educational levels range from **elementary** to **competition**. This places the dataset between two established traditions. On one side are multimodal STEM assessment datasets such as the benchmark in "Measuring Vision-Language STEM Skills of Neural Models," which emphasizes K–12 vision-language multiple-choice questions with train/valid/test splits [2402.17205]. On the other side are formal logical reasoning resources, such as automatically generated first-order logic problems in Zermelo–Fraenkel set theory [2502.14180] and synthetic rule-learning corpora in Datalog form [1909.07095]. Logics-STEM-SFT is distinct in being a large-scale **textual reasoning corpus with distilled chain-of-thought**, rather than a pure benchmark or a narrowly formal-logic generator.

A common misconception is that the dataset is simply a benchmark under the Logics-STEM name. The paper instead defines it as a release for training, with **all downstream evaluation done on external STEM benchmarks** rather than on an internal held-out split [2601.01562].

## 2. Scale, composition, and record structure

The release is organized around two scales and a fixed record schema.

| Release | Scale | Notes |
|---|---:|---|
| Logics-STEM-SFT-10M | ~10 million | Full training corpus |
| Logics-STEM-SFT-2.2M | 2.2 million | Stratified downsample |
| 2.2M subset composition | ~1.05 M / ~1.14 M | Pure mathematics / broader STEM |

Each example contains the following fields: the original **question**, a **chain_of_thought** consisting of a multi-step, teacher-distilled reasoning trace ending in a boxed final answer, the extracted **answer**, and a **meta** object containing annotations for **domain**, **educational level**, and **answer type**. The release format is **compressed JSONL**, sharded for parallel I/O, with **index files provided for random access** [2601.01562].

The paper gives a schema example of the form:

```json
{
  "id": "e4a1f22d5c6b3a7d…",
  "source": "NuminaMath-1.5",
  "question": "...",
  "meta": {
    "domain": "Math",
    "level": "competition",
    "answer_type": "numeric"
  },
  "chain_of_thought": "...step-by-step solution...",
  "answer": "42"
}
```

The metadata taxonomy is central to the dataset’s reuse. Domains distinguish mathematics from broader STEM; educational level ranges over **elementary, junior-secondary, senior-secondary, undergraduate, graduate, competition**; and answer types include **Boolean, multiple-choice, numeric, vector/matrix, interval, expression, string, proof, explanation, other**. The full corpus therefore supports stratification by both disciplinary content and response modality [2601.01562].

Several distributional statistics are reported. After deduplication and decontamination, the corpus size is approximately **10 M**. In the 2.2M subset there are approximately **1.05 M pure mathematics samples** and **1.14 M broader-STEM samples**. Chain-of-thought lengths range from **10–200+ tokens**, with **median ~50**, and the average prompt-plus-CoT context is **~600 tokens**. The paper further states that **lexical diversity (type–token ratio) is high**, reflecting multi-disciplinary vocabulary [2601.01562].

## 3. Five-stage curation pipeline

The dataset is defined operationally by a five-stage curation engine.

**Stage 1: Annotation.** Raw questions are processed by **Qwen3-235B-Instruct** to annotate validity or unambiguity, discipline, educational level, answer type, and verifiable answer status. Questions that are **unsolvable, incomplete, or missing-image items** are filtered out. The discipline labeling includes **6 buckets: Math; STEM w/o Math; Humanities/Social Science; Business; Medicine/Biology; Code**. Invalid or ambiguous questions are discarded [2601.01562].

**Stage 2: Deduplication.** Exact duplicates are removed using **MD5(question)** fingerprints. Near duplicates are removed using **MinHash with 24 bands of width 10 to approximate Jaccard**. Within each MinHash bucket, only one example is kept, with priority given to **samples with verifiable answers**. This combination is designed to reduce both literal repetition and near-paraphrastic overlap [2601.01562].

**Stage 3: Decontamination.** Training samples are removed if they either **share a MinHash bucket** with any held-out evaluation example or **contain an identical 13-gram** with such an example. The paper describes this as a dual check that drives overlap to **near zero** [2601.01562].

**Stage 4: Response Distillation.** Chain-of-thought traces are generated by **Qwen3-235B-A22B-Thinking-2507** using the configuration **temperature = 0.6, top_k = 20, top_p = 0.95, max_context_length = 32768**. CoTs whose **n-gram duplication ratio exceeds a preset threshold** are discarded. The paper notes that **optional answer-verification via math-verify** is available, but also states that **erroneous CoTs are retained in small quantities since aggressive filtering was shown to degrade model robustness** [2601.01562].

**Stage 5: Stratified Sampling.** The proxy for difficulty is the token length $l$ of the distilled CoT. Let $Q_p$ denote the $p$-th percentile of the $\{l_i\}$. The sampling weight is defined as

$$
w(l)=
\begin{cases}
1.0, & l\ge Q_{0.75},\\
0.5, & Q_{0.50}\le l<Q_{0.75},\\
0.1, & Q_{0.20}\le l<Q_{0.50},\\
0.0, & l<Q_{0.20}.
\end{cases}
$$

The 2.2M release is then sampled with probability $P_i \propto w(l_i)$ [2601.01562].

This pipeline indicates that “quality” is operationalized not only as correctness and cleanliness, but also as diversity, benchmark isolation, and controlled retention of harder reasoning traces.

## 4. Quality controls, release conventions, and what the dataset is not

The release provides a training corpus **without internal val/test splits**. The paper is explicit that **all downstream evaluation is done on external STEM benchmarks**, including **AIME2024/5, HMMT2025, BeyondAIME, GPQA-Diamond, MMLU-Pro-STEM, CMMLU-STEM, R-Bench, etc.** [2601.01562].

Quality measurements are tied to the pipeline stages. Annotation quality is enforced by the **Instruct teacher model’s built-in heuristics**. Deduplication and decontamination coverage are measured by **bucket-drop rates (≈67% dropped to reach 10 M)**. Distillation quality is monitored using **n-gram repetition ratios** and **math-verify answer-match rates** [2601.01562].

Several misunderstandings are corrected by these release conventions. First, the dataset is not a multimodal resource: the paper states that the **question** field contains the original STEM problem with **images excluded**. This contrasts with the earlier STEM benchmark, where every retained example has at least one image and the schema includes `context_image` plus possibly image-based answer options [2402.17205]. Second, the dataset is not described as fully verified ground truth CoT. The paper explicitly retains a small quantity of erroneous CoTs because aggressive filtering harmed robustness [2601.01562].

A plausible implication is that the dataset is optimized for post-training robustness rather than for creating a clean symbolic proof corpus in the formal-methods sense.

## 5. Role in failure-driven post-training

Logics-STEM-SFT-Dataset is embedded in a broader **data-algorithm co-design** framework. The abstract attributes the model gains to joint optimization of data and algorithm to fit a “gold-standard distribution behind reasoning,” and the detailed description formalizes this with a first-stage SFT objective, a failure indicator, a failure-biased prompt distribution, a retrieval kernel, a synthesized training distribution, and a final mixture distribution [2601.01562].

The first-stage SFT objective is

$$
\theta_1
= \arg\min_{\theta}
\mathbb{E}_{(x,y)\sim P_0}\Biggl[-\sum_{t=1}^T\log\pi_\theta\!\bigl(y^t\mid x,y^{<t}\bigr)\Biggr].
$$

The binary failure indicator is

$$
w_{\theta_1}(x)
= \mathbf{1}\{\text{model’s final answer on }x\text{ is incorrect}\}.
$$

This induces a failure-biased prompt distribution

$$
Q_{\theta_1}(x)
= \frac{Q(x)\,w_{\theta_1}(x)}
{\mathbb{E}_{x\sim Q}[\,w_{\theta_1}(x)\,]}.
$$

Document retrieval is modeled with a top-$k$ kernel

$$
K_{\tau,k}(d\mid x)
\propto \exp\!\Bigl(\langle e(x),e(d)\rangle/\tau\Bigr)\,
\mathbf{1}[d\in\mathrm{Top}\!-\!k(x)].
$$

The synthesized training distribution is

$$
P_{\mathrm{syn}}(x',y') = \sum_x Q_{\theta_1}(x)\sum_d K_{\tau,k}(d\mid x)\,
G_{\mathrm{acc}}(x',y'\mid d).
$$

The final second-stage objective uses

$$
P_1 = \lambda\,P_0 \;+\;(1-\lambda)\,P_{\mathrm{syn}},
\quad
\theta_2=\arg\min_\theta\mathbb{E}_{(x,y)\sim P_1}\bigl[\ell_\theta(x,y)\bigr].
$$

Within this formalism, Logics-STEM-SFT supplies the initial supervised distribution $P_0$ and, by construction, the metadata needed for targeted sampling and failure analysis. The dataset is therefore not merely a corpus of solved problems; it is part of a feedback loop in which failures trigger retrieval and synthesis around failure regions [2601.01562].

## 6. Empirical effects and relation to adjacent dataset traditions

The paper reports that fine-tuning **Qwen3-8B** on **Logics-STEM-SFT-2.2M** yields measurable gains on multiple external reasoning benchmarks: **AIME2024 @Pass@1: 76.0% → 80.62%**, **AIME2025: 67.3% → 73.33%**, **GPQA-Diamond: 62.0% → 72.70%**, and **MMLU-Pro-STEM: 83.02% → 85.20%** [2601.01562]. These results locate the dataset within contemporary post-training practice: it is intended to change model behavior on out-of-distribution evaluations rather than to serve as a closed-form testbed.

In the wider research landscape, Logics-STEM-SFT occupies a different niche from at least three adjacent resources. The multimodal STEM benchmark of 1,073,146 questions spans **448 skills** and emphasizes **vision-language** K–12 evaluation with fixed train/valid/test partitions [2402.17205]. The first-order logic datasets generated in Zermelo–Fraenkel set theory emphasize truth/falsity judgments on prenex formulas whose difficulty is controlled by graph parameters and negation structure [2502.14180]. RuDaS, by contrast, generates synthetic Datalog rule-learning corpora and evaluates learned rules using **Herbrand-based measures**, **IR measures**, and **R-score** [1909.07095].

This comparison clarifies what “Logics” means in Logics-STEM-SFT. The paper does not define the dataset as a purely formal-logic corpus. Instead, it combines **large-scale open-source data** with **carefully designed synthetic data** and distills long reasoning traces across mathematics and broader STEM [2601.01562]. A plausible implication is that the dataset’s “logical” character lies in post-training for multi-step reasoning, not in restricting the corpus to symbolic logic instances.

From an encyclopedic perspective, the defining properties of Logics-STEM-SFT-Dataset are therefore its **10M-scale text-only reasoning triples**, its **five-stage curation pipeline**, its **metadata-rich stratified downsampling**, and its explicit integration into **failure-driven post-training** for STEM reasoning models [2601.01562].

Source: https://www.emergentmind.com/topics/logics-stem-sft-dataset