Get the app
ablate 0 0 0

PersonaBias/alpha_beta_intervention

published by @PersonaBias

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
196
Last 30 days
130
Likes
-
Size
2.7 MB
Created on HF
2026-09-15
Age
1 months ago

Tags

task_categories:text-classificationtask_categories:question-answeringlicense:mitsize_categories:10K<n<100Kformat:csvmodality:tabularmodality:textlibrary:datasetslibrary:pandaslibrary:polarslibrary:mlcroissantregion:usbiaspersonamechanistic-interpretabilityactivation-steeringalpha-beta-intervention

README current version from Hugging Face

license: mit
task_categories:

  • text-classification
  • question-answering
    tags:
  • bias
  • persona
  • mechanistic-interpretability
  • activation-steering
  • alpha-beta-intervention
    pretty_name: "Persona Bias -- AHI Alpha/Beta Intervention (per head set)"
    size_categories:
  • 10K<n<100K
    configs:
  • config_name: "Llama-3.1-8B-Instruct__arc_challenge__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_challenge/emotion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__arc_challenge__race__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_challenge/race/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__arc_challenge__religion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_challenge/religion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__arc_easy__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_easy/emotion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__arc_easy__gender__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_easy/gender/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__arc_easy__race__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_easy/race/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__arc_easy__religion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/arc_easy/religion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__ethics__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/ethics/emotion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__ethics__gender__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/ethics/gender/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__ethics__race__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/ethics/race/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__ethics__religion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/ethics/religion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__safety__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/safety/emotion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__safety__gender__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/safety/gender/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__safety__race__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/safety/race/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__safety__religion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/safety/religion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__sst2__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/sst2/emotion/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__sst2__gender__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/sst2/gender/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__sst2__race__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/sst2/race/ahi_rates.csv
  • config_name: "Llama-3.1-8B-Instruct__sst2__religion__rates"
    data_files:
    • split: train
      path: Llama-3.1-8B-Instruct/sst2/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_challenge__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_challenge/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__arc_easy__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/arc_easy/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__ethics__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/ethics/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__safety__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/safety/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__emotion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/emotion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__gender__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/gender/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__race__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/race/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/religion/ahi_rates.csv
  • config_name: "Llama-3.2-3B-Instruct__sst2__religion__rates"
    data_files:
    • split: train
      path: Llama-3.2-3B-Instruct/sst2/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_challenge__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_challenge/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_easy__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_easy/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_easy__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_easy/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_easy__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_easy/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_easy__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_easy/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_easy__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_easy/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__arc_easy__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/arc_easy/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__ethics__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/ethics/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__safety__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/safety/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__emotion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/emotion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__gender__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/gender/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__race__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/race/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/religion/ahi_rates.csv
  • config_name: "Qwen2.5-7B-Instruct__sst2__religion__rates"
    data_files:
    • split: train
      path: Qwen2.5-7B-Instruct/sst2/religion/ahi_rates.csv
  • config_name: "summary__rates_all"
    data_files:
    • split: train
      path: summary/rates_all.csv

README history 3 revisions

Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.

  1. 2026-09-22Upload 60 files5a86e9015 KB
    Loading...
  2. 2026-09-15Upload 41 files079914612.2 KB
    Loading...
  3. 2026-09-15initial commite40d9a50 B
    Loading...

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Community discussions 1 thread · 1 open · 1 comments

Threads posted by the community on Hugging Face: bug reports, questions about usage, pull requests on the README, license clarifications. Click any thread to expand and read the full conversation inline. Newest first.

  1. 2026-09-16[bot] Conversion to Parquetopen1 💬#1
    Loading...

Continue on Hugging Face →

Related datasets same stage · ablate

lasrprobegen/refusal-activations
74.6K dl 1 likes 1.21 TB
Refusal Activations Dataset This dataset is now configured to load the full ~97k samples from jailbreak_mixed_100k.csv.
lu-christina/assistant-axis-vectors
14.4K dl 13 likes 1.2 GB
The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models This repository contains pre-computed axes...
stablellama/erotic-image-prompts
11.9K dl 66 likes 104.7 MB
Dataset Card for Erotic Image Prompts Dataset Description Dataset Summary Large language models (...
pandaman007/assistant-axis-abliteration-vectors
9.5K dl 24.5 GB
pandaman007/steered-persona-space
9.1K dl 539.6 MB
Steered Persona Space How activation steering reshapes a language model's character cloud. We apply activation-steering hooks a...
Undi95/orthogonal-activation-steering-TOXIC
7.2K dl 19 likes 2.1 MB
sunnycloudhust/CHaRS-activations
5.3K dl 2 likes 63.8 GB
CHaRS: Concept Heterogeneity-Aware Representation Steering This repository contains implementation of CHaRS (Concept Heterogene...
dementor-research/dementor-complete-experiment-results
5.1K dl 8.2 GB
Dementor complete experiment results Audited outputs for the configuration-defined Dementor completion campaign. Au...

Read further

Community best practices (ablation)
5 rules from KL-optimisation to layer-0 preservation.
M1 - Direct removal
The canonical Arditi-style ablation.
M3 - Layer-wise ablation
Selective abliteration limited to a band of layers.
Non-English abliteration communities
The Chinese layer-window heuristic and huihui-ai practice.

Metadata

License
mit

Download the dataset

0

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python) 
from datasets import load_dataset
ds = load_dataset("PersonaBias/alpha_beta_intervention")
Raw snapshot (Python) 
from huggingface_hub import snapshot_download
snapshot_download(repo_id="PersonaBias/alpha_beta_intervention", repo_type="dataset")
git clone (requires git-lfs) 
git clone https://huggingface.co/datasets/PersonaBias/alpha_beta_intervention
Browse repository on Hugging Face ->
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Get the app Browse the catalog
Open in Abliteration