lugman/LFM2.5-350M-SmolDataAgent
Text Generation • 0.4B • Updated • 61
task_id stringlengths 20 23 | difficulty_tier stringclasses 3
values | source stringclasses 6
values | tools listlengths 1 1 | messages listlengths 7 26 |
|---|---|---|---|---|
0000_369_369503_qa_1 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0000_641_641256_qa_1 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0000_656_656399_qa_2 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0000_806_806826_qa_3 | medium | synthetic:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0000_981_981197_qa_1 | medium | haiku:G3_analysis | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0000_982_982280_qa_2 | easy | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_074_1074738_qa_1 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_085_1085629_qa_2 | hard | haiku:G3_analysis | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_137_1137361_qa_2 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_155_1155264_qa_5 | medium | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_170_1170198_qa_3 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_175_1175291_qa_4 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_182_1182948_qa_1 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_188_1188925_qa_3 | easy | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_189_1189227_qa_1 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_189_1189227_qa_2 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_191_1191057_qa_4 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_196_1196803_qa_3 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_197_1197721_qa_2 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_231_1231918_qa_2 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_234_1234901_qa_3 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_234_1234901_qa_5 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_238_1238370_qa_1 | medium | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_239_1239559_qa_4 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_240_1240535_qa_1 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_243_1243037_qa_1 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_244_1244861_qa_4 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_257_1257061_qa_1 | easy | synthetic:G1_lookup | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_257_1257756_qa_2 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_257_1257756_qa_3 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
0001_261_1261978_qa_5 | medium | lfm:G2_multistep | [
{
"type": "function",
"function": {
"name": "bash",
"description": "Run a shell command in the sandbox and return its combined stdout+stderr. The shell is non-stateful between calls (each call is a fresh `bash -c`). Use it to explore files (ls, head, cat), run Python (`python3 -c ...`), and writ... | [
{
"role": "system",
"content": "You are an autonomous data-analysis agent operating in a sandboxed Linux container. Your only tool is `bash`. The dataset files are in /home/user/input/. Python 3 + pandas + numpy + scikit-learn + scipy are pre-installed.\n\nWork the problem step-by-step: first inspect the da... |
End of preview. Expand in Data Studio
SmolDataAgent-Trajectories
2,833 multi-turn agent trajectories for training small data-analysis agents, built on tasks from SmolDataEnvs. Used to train lugman/LFM2.5-350M-SmolDataAgent.
In each trajectory the agent answers a question about a sandbox of data files using a single bash tool, and writes its final answer to /workdir/answer.txt.
Composition
The trajectories started as rollouts from two teacher models, LFM2.5-2.6B and Claude Haiku 5.5. They were then edited (cleaned, and trimmed of tool calls made after the answer was written) and synthetically enriched.
| Source | easy | medium | hard | Total |
|---|---|---|---|---|
| LFM2.5-2.6B | 91 | 1,267 | 21 | 1,379 |
| Claude Haiku 5.5 | 40 | 140 | 140 | 320 |
| Synthetic enrichment | 1,007 | 126 | 1 | 1,134 |
| Total | 1,138 | 1,533 | 162 | 2,833 |
| Source | G1 lookup | G2 multistep | G3 analysis |
|---|---|---|---|
| LFM2.5-2.6B | 0 | 1,362 | 17 |
| Claude Haiku 5.5 | 0 | 0 | 320 |
| Synthetic enrichment | 1,112 | 19 | 3 |
Why not SmolDataEnvs-sft
FineEnvs/SmolDataEnvs-sft is the existing SFT set for this environment. A custom set was generated instead because:
- Mixed harnesses. Its rows come from two agent harnesses (
bash,seta) with different habits, e.g. 71% ofsetarows use a "Plan:" scaffold. Every trajectory here follows one convention. - Lenient grading. Its grader takes the first number anywhere in the answer; 10.4% of re-gradable rows fail the stricter single-value grader used here.
- Better results. In a controlled comparison, LFM2.5-350M trained on this custom curated set outperformed the same model trained on SmolDataEnvs-sft.
Format
One JSON object per line in train.jsonl:
| Field | Description |
|---|---|
task_id |
SmolDataEnvs task identifier |
difficulty_tier |
easy, medium or hard |
source |
<origin>:<group>, where origin is lfm (LFM2.5-2.6B), haiku (Claude Haiku 5.5) or synthetic, and group is G1_lookup, G2_multistep or G3_analysis |
tools |
OpenAI-style tool definitions (the bash tool) |
messages |
Chat messages (system, user, assistant with tool_calls, tool) in OpenAI/TRL format |
from datasets import load_dataset
ds = load_dataset("lugman/SmolDataAgent-Trajectories", split="train")
messages = ds[0]["messages"]
- Downloads last month
- 56
Models trained or fine-tuned on lugman/SmolDataAgent-Trajectories