metadata
license: agpl-3.0
tags:
- smoltrace
- smolagents
- evaluation
- benchmark
- llm
- agents
Tiny Agents. Total Visibility.
SMOLTRACE Evaluation Results
This dataset contains evaluation results from a SMOLTRACE benchmark run.
Dataset Information
| Field | Value |
|---|---|
| Model | openai/gpt-oss-120b |
| Run ID | job_4acee6f5 |
| Agent Type | both |
| Total Tests | 15 |
| Generated | 2025-12-06 04:13:45 UTC |
| Source Dataset | kshitijthakkar/smoltrace-tasks |
Schema
| Column | Type | Description |
|---|---|---|
model |
string | Model identifier |
evaluation_date |
string | ISO timestamp of evaluation |
task_id |
string | Unique test case identifier |
agent_type |
string | "tool" or "code" agent type |
difficulty |
string | Test difficulty level |
prompt |
string | Test prompt/question |
success |
bool | Whether the test passed |
tool_called |
bool | Whether a tool was invoked |
correct_tool |
bool | Whether the correct tool was used |
final_answer_called |
bool | Whether final_answer was called |
response_correct |
bool | Whether the response was correct |
tools_used |
string | Comma-separated list of tools used |
steps |
int | Number of agent steps taken |
response |
string | Agent's final response |
error |
string | Error message if failed |
trace_id |
string | OpenTelemetry trace ID |
execution_time_ms |
float | Execution time in milliseconds |
total_tokens |
int | Total tokens consumed |
cost_usd |
float | API cost in USD |
enhanced_trace_info |
string | JSON with detailed trace data |
Usage
from datasets import load_dataset
# Load the results dataset
ds = load_dataset("YOUR_USERNAME/smoltrace-results-TIMESTAMP")
# Filter successful tests
successful = ds.filter(lambda x: x['success'])
# Calculate success rate
success_rate = sum(1 for r in ds['train'] if r['success']) / len(ds['train']) * 100
print(f"Success Rate: {success_rate:.2f}%")
Related Datasets
This evaluation run also generated:
- Traces Dataset: Detailed OpenTelemetry execution traces
- Metrics Dataset: GPU utilization and environmental metrics
- Leaderboard: Aggregated metrics for model comparison
About SMOLTRACE
SMOLTRACE is a comprehensive benchmarking and evaluation framework for Smolagents - HuggingFace's lightweight agent library.
Key Features
- Automated agent evaluation with customizable test cases
- OpenTelemetry-based tracing for detailed execution insights
- GPU metrics collection (utilization, memory, temperature, power)
- CO2 emissions and power cost tracking
- Leaderboard aggregation and comparison
Quick Links
Installation
pip install smoltrace
Citation
If you use SMOLTRACE in your research, please cite:
@software{smoltrace,
title = {SMOLTRACE: Benchmarking Framework for Smolagents},
author = {Thakkar, Kshitij},
url = {https://github.com/Mandark-droid/SMOLTRACE},
year = {2025}
}
Generated by SMOLTRACE