Datasets:
id stringlengths 13 13 | question stringlengths 212 4.16k | response stringlengths 220 410k | num_messages int64 3 135 | messages listlengths 3 135 |
|---|---|---|---|---|
stage1-000000 | A standard potassium poly(vinylsulfate) solution contains PVSK polymer chains at concentration \(0.002500\ \mathrm{mol\,dm^{-3}}\). Each PVSK chain has exactly 40 anionic sulfate charge sites, and each anionic charge site neutralizes exactly one benzyldimethyltetradecylammonium cation. Thus, the stated PVSK concentrati... | <think>
The calculation confirms x = 0.25. Let me present the full solution.</think><answer>
## Solution
**Step 1 — Correct charge-site concentration of the PVSK standard**
The stated concentration is the *chain* concentration. Since each chain bears 40 sulfate sites:
\[
c_{\text{sites}} = 0.002500 \times 40 = 0.100... | 5 | [
{
"role": "system",
"content": "You are a deep research assistant. Your core function is to conduct thorough, multi-source investigations into any topic. You must handle both broad, open-domain inquiries and queries within specialized academic fields. For every request, synthesize information from credible,... |
stage1-000001 | A singly charged protein ion is accelerated through a potential \(V\) and then traverses a field-free flight tube of length \(L\). In the nonrelativistic model,
\[
\tau(M)=L\sqrt{\frac{M}{2eVN_A}},
\]
where \(M\) is the molar mass, \(e\) is the elementary charge, and \(N_A\) is Avogadro’s constant. A single observed fl... | <think>
We need answer math. Need parse. Need derive. We can use python but likely symbolic. Need produce closed form S.
We need be careful. Problem: singly charged protein ion accelerated through V, flight tube length L. Nonrelativistic flight time tau(M)=L sqrt(M/(2 e V N_A)). Observed t with Gaussian timing uncerta... | 3 | [
{
"role": "system",
"content": "You are a deep research assistant. Your core function is to conduct thorough, multi-source investigations into any topic. You must handle both broad, open-domain inquiries and queries within specialized academic fields. For every request, synthesize information from credible,... |
stage1-000002 | In colloid science, use the following standard facts: lyophobic colloids are not obtained merely by simple mixing of the dispersed phase with the dispersion medium; an emulsion is a colloidal system in which both the dispersed phase and the dispersion medium are liquids; micelles form only in suitable surfactant–solven... | <think>
We need answer. Need parse. Standard facts tell P1 true, P2 true, P3 false, P4 false. Thus T={1,2}. Need determine set of Coxeter lengths of permutations in S5 whose right descent set exactly {1,2}. Need list increasing.
Need be careful definition right descent set D_R(w) = {i: w_i > w_{i+1}} in one-line notat... | 3 | [
{
"role": "system",
"content": "You are a deep research assistant. Your core function is to conduct thorough, multi-source investigations into any topic. You must handle both broad, open-domain inquiries and queries within specialized academic fields. For every request, synthesize information from credible,... |
stage1-000003 | For a fixed positive integer \(k\), consider an idealized non-linear molecule with full \(I_h\) symmetry formed from \(k\) concentric, similar \(\mathrm{C}_{60}\)-type shells at distinct radii, so that there are \(60k\) atoms. Assume the full Cartesian displacement representation is the direct sum of \(k\) copies of th... | <think>
We need answer. Need be careful. Let's parse.
We have idealized nonlinear molecule with full I_h symmetry formed from k concentric similar C60-type shells distinct radii, 60k atoms. Full Cartesian displacement representation is direct sum of k copies of one-shell representation Γ_shell given:
2A_g ⊕ A_u ⊕ 4T_{... | 3 | [
{
"role": "system",
"content": "You are a deep research assistant. Your core function is to conduct thorough, multi-source investigations into any topic. You must handle both broad, open-domain inquiries and queries within specialized academic fields. For every request, synthesize information from credible,... |
stage1-000004 | Let \(n\ge 2\) be an integer, let \(R>0\), and define
\[
A_k=R e^{2\pi i(k-1)/n},\qquad k=1,\dots,n.
\]
For \(t>0\), set
\[
E_t=\left\{z\in\mathbb C:\prod_{k=1}^n |z-A_k|\le t\cdot \max_{|w|=R}\prod_{k=1}^n |w-A_k|\right\}.
\]
Let \(S_{n,R}(t)\) be the least real number \(S\ge 0\) such that
\[
E_t\subseteq \{z\in\mathb... | <think>
We need answer math. Need determine S_{n,R}(t) closed form. Need be careful. We have n roots of unity radius R. E_t lemniscate sublevel of product polynomial P(z)=∏(z-A_k)=z^n-R^n. max on |w|=R of |w^n-R^n|. Compute max. For w=R e^{iθ}, |R^n e^{inθ}-R^n|=R^n |e^{inθ}-1|=R^n * 2 |sin(nθ/2)|. Max = 2 R^n. Thus E_... | 3 | [
{
"role": "system",
"content": "You are a deep research assistant. Your core function is to conduct thorough, multi-source investigations into any topic. You must handle both broad, open-domain inquiries and queries within specialized academic fields. For every request, synthesize information from credible,... |
stage1-000005 | Let \(0<\varepsilon<\pi\), and on the punctured disc \(D^*=\{z\in\mathbb C:0<|z|<\varepsilon\}\) consider the rank-one meromorphic connection whose horizontal sections \(y\) satisfy
\[
\frac{dy}{dz}=\frac{1}{z^2\sin z}\,y.
\]
Fix the basepoint \(z_0=\varepsilon/2\), choose a nonzero local horizontal section \(y\) near ... | <think>
We need answer. Need solve math. Let's restate.
Connection dy/dz = (1/(z^2 sin z)) y on punctured disc radius eps < pi. So singular at z=0. Horizontal section y = exp(∫ 1/(z^2 sin z) dz) local. Monodromy around loop counterclockwise. F = z^{-1/3} y^2. Need scalar multiplier Q after analytic continuation once C... | 3 | [
{
"role": "system",
"content": "You are a deep research assistant. Your core function is to conduct thorough, multi-source investigations into any topic. You must handle both broad, open-domain inquiries and queries within specialized academic fields. For every request, synthesize information from credible,... |
stage1-000006 | "A two-nucleon strong interaction is to be modeled from the following self-contained empirical const(...TRUNCATED) | "<think>\nWe need answer. Need solve. Need determine interaction class from constraints. Then comput(...TRUNCATED) | 3 | [{"role":"system","content":"You are a deep research assistant. Your core function is to conduct tho(...TRUNCATED) |
stage1-000007 | "Let \\(e,n_i,\\mu_n,\\mu_p>0\\) with \\(\\mu_n\\neq \\mu_p\\). For each \\(n>0\\), impose the mass-(...TRUNCATED) | "<think>\nWe need answer. Need think carefully. We have measure dP_beta(n)=1/Z exp(-beta sigma(n)) d(...TRUNCATED) | 3 | [{"role":"system","content":"You are a deep research assistant. Your core function is to conduct tho(...TRUNCATED) |
stage1-000008 | "At a fixed temperature, liquids \\(A\\) and \\(B\\) form an ideal binary solution over the full com(...TRUNCATED) | "<think>\nWe need answer. Need parse. Need compute intrinsic Laplace-Beltrami of P on simplex with m(...TRUNCATED) | 3 | [{"role":"system","content":"You are a deep research assistant. Your core function is to conduct tho(...TRUNCATED) |
stage1-000009 | "Let \\(A\\) be the integer matrix whose rows record conservation of Fe, O, H, and electric charge, (...TRUNCATED) | "<think>\nWe need answer. Need solve math. We need compute lattice ideal quotient Hilbert series lim(...TRUNCATED) | 3 | [{"role":"system","content":"You are a deep research assistant. Your core function is to conduct tho(...TRUNCATED) |
End of preview. Expand in Data Studio
Axion 1.0 Training Data
Scientific reasoning and tool use for open agent foundation models.
Project Page · Paper · Code
Axion 1.0 follows a three-stage training recipe: learn self-contained reasoning, learn to acquire missing information through tools, then refine these abilities through reinforcement learning. This repository contains the two SFT trajectory splits and the RL training tasks used by the 122B R3 run.
Dataset at a glance
| Configuration | Split | Records | Contents |
|---|---|---|---|
default |
stage1 |
10,819 | Self-contained scientific reasoning; Python execution, without retrieval |
default |
stage2 |
7,544 | Verified tool-use trajectories for tasks constructed from public databases |
rl |
train |
484 | RL prompts, reference answers, and task metadata |
The two SFT splits contain 18,363 trajectories. Each row is a complete
conversation, including its system prompt, reasoning, tool calls, tool responses,
and assistant messages. Tool responses are represented within user messages in
the released conversation format.
Data construction
Stage I — learn to reason. Model-aware construction strengthens self-contained scientific problems. Solvers can use Python to carry out computations.
Stage II — learn to investigate. Tasks require external information, with reference answers established through structured data and deterministic computation. The released records contain the collected solution trajectories.
See the paper for the construction, verification, training, and evaluation procedures.
Browse the trajectories
Select the default configuration and the stage1 or stage2 split in the
Dataset Viewer. The table presents the question and the last assistant response
alongside the complete conversation.
| Column | Description |
|---|---|
id |
Stable identifier based on the stage and original row position |
question |
First user message, copied from the conversation |
response |
Last assistant message, retaining its original reasoning and answer tags |
num_messages |
Number of messages, including the system prompt and tool-response messages |
messages |
Complete ordered list of {role, content} objects |
question and response are browsing columns derived from messages. Use
messages to train on the complete multi-turn trajectory.
Load the data
from datasets import load_dataset
repo = "Axion1-0/axion-training-data"
stage1 = load_dataset(repo, name="default", split="stage1")
stage2 = load_dataset(repo, name="default", split="stage2")
rl_tasks = load_dataset(repo, name="rl", split="train")
conversation = stage2[0]["messages"]
The RL configuration has a separate schema: prompt, reward_model,
rl_task_category, data_source, and extra_info. It contains tasks and reference
answers, rather than completed SFT solution trajectories.
Files
assets/ Logo and data-construction figure
data/stage1-*.parquet Stage I SFT trajectories
data/stage2-*.parquet Stage II SFT trajectories
rl.parquet RL training tasks
The repository distributes the complete SFT conversations and RL tasks in Parquet format. The Dataset Viewer and loading examples read these files directly.
Environment-specific runtime path prefixes in a small number of tool outputs have been normalized.
Reference
Axion 1.0: Open Agent Foundation Models, Simplified
Yuqi Jia, Yunjie Tian, Jiayu Zheng, Yijun Liang, Jingchen Sun, Mengyu Zhao, and Di Fu.
License
This dataset is licensed under the Apache License 2.0.
- Downloads last month
- -