license: mit
task_categories:

• text-classification
  language:
• en
  tags:
• ai-safety
• alignment
• benchmarks
• llm-evaluation
• red-teaming
• jailbreak
• safety
  pretty_name: Sentinel Evaluations
  size_categories:
• n<1K
  configs:
• config_name: seeds
  data_files:
  • split: train
    path: seeds.jsonl
• config_name: results
  data_files:
  • split: train
    path: results.jsonl

Sentinel Evaluations

Evaluation results for multiple alignment seeds across various AI safety benchmarks.

Overview

This dataset contains:

• Seeds: Alignment prompts from different sources (Sentinel, FAS, Safyte xAI)
• Results: Evaluation results across HarmBench, JailbreakBench, GDS-12, and more

Quick Start

from datasets import load_dataset

# Load seeds
seeds = load_dataset("sentinelseed/sentinel-evaluations", "seeds", split="train")

# Load results
results = load_dataset("sentinelseed/sentinel-evaluations", "results", split="train")

# Filter by seed
sentinel_results = results.filter(lambda x: x["seed_id"] == "sentinel-v2")

# Filter by benchmark
harmbench_results = results.filter(lambda x: x["benchmark_id"] == "harmbench")

Seeds Included

Benchmarks

Results Schema

{
    "seed_id": "sentinel-v2",
    "benchmark_id": "harmbench",
    "model": "gpt-4o",
    "date": "20251208",
    "total_tests": 200,
    "safe_responses": 196,
    "unsafe_responses": 4,
    "safety_rate": 0.98,
    "source": "sentinel-platform"
}

Summary Results

Sentinel v2 THSP

Safyte xAI

Links

• Website: sentinelseed.dev
• Lab: sentinelseed.dev/evaluations
• Seeds Dataset: sentinelseed/alignment-seeds
• GitHub: github.com/sentinel-seed
• Contact: [email protected]

Citation

@misc{sentinel2025,
  title={Sentinel Seed: Practical AI Alignment for Developers},
  author={Sentinel Team},
  year={2025},
  url={https://sentinelseed.dev}
}

License

MIT License - Sentinel Team | [email protected]