玩模型开放模型 · 清晰来源 · 自主选择
返回模型目录
ai-safety-institute /

apollo-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organism-ge-ae501e94

HF 源站

ai-safety-institute/apollo-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organism-ge-ae501e94 is an uncensored open-source model hosted on Hugging Face.

PyTorchLicense: mit27Bdeception-detectionprobeapollo
README.md说明 / 使用方法

tags:

  • deception-detection
  • probe
  • apollo library_name: pytorch license: mit

Apollo-Style Deception Probe for mlabonne/gemma-3-27b-it-abliterated:aletheias-quest/collusion-model-organism-gemma3-27b-v1

A probe trained to detect deceptive behaviour in mlabonne/gemma-3-27b-it-abliterated:aletheias-quest/collusion-model-organism-gemma3-27b-v1 using residual stream activations, following the methodology from Detecting Strategic Deception in Language Models (Apollo Research, 2024).

Quick Start

uv add lie-detectors        # or: pip install lie-detectors
from lie_detectors import get_probe

probe = get_probe("ai-safety-institute/apollo-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organism-ge-ae501e94")

The default checkpoint is the best performer from the hyperparameter sweep (l_49_lm_500000_ar_lr.pt). To pick a specific checkpoint, pass filename=:

probe = get_probe("ai-safety-institute/apollo-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organism-ge-ae501e94", filename="l_40_ar_mlp_wd_0_001_lr_0_0001_ep_100.pt")

See UKGovernmentBEIS/lie_detectors for the loading library.

Use sweep.json to see all 296 available checkpoints and their metrics.

Model Details

PropertyValue
Target modelmlabonne/gemma-3-27b-it-abliterated:aletheias-quest/collusion-model-organism-gemma3-27b-v1
Safe namemlabonne--gemma-3-27b-it-abliterated@aletheias-quest--collusion-model-organism-gemma3-27b-v1
Default checkpointl_49_lm_500000_ar_lr.pt
Available checkpoints296
CalibrationThreshold set at 1% FPR on Alpaca (honest baseline)

Training Data

Probes are trained on an instructed pairs dataset (model instructed to be deceptive vs. honest) based on Facts True False and calibrated on Alpaca (honest-only baseline) to achieve a 1% false positive rate.

Citation

Original Paper

@misc{goldowskydill2025detectingstrategicdeceptionusing,
      title={Detecting Strategic Deception Using Linear Probes},
      author={Nicholas Goldowsky-Dill and Bilal Chughtai and Stefan Heimersheim and Marius Hobbhahn},
      year={2025},
      eprint={2502.03407},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2502.03407},
}

Trained Probes

@misc{cooney2026liedetectors,
      title={``Did you lie?'' Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms},
      author={Alan Cooney and David Africa and Geoffrey Irving},
      year={2026},
      month={May},
}
HUGGING FACE

上游数据

HF 下载量0
HF 点赞量0

HF 下载量为源站最近 30 天统计,与本站点击分开。

SITE STATS

本站统计

下载点击—
请求数—
匿名去重—最近 7 天—

本站下载量记录按钮点击,不代表下载完成;短时连点会去重。

来源与版本

作者
ai-safety-institute
最近更新
2026/06/25
最新创建
2026/06/25
HF 趋势
未知
数据快照
2026/09/17
版本
未知
许可证
mit