CoolFace
Modelpublic

SallySims/equibert-awareness-scorer

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face
0likes10downloads
Model Card

EquiBERT — DEI Awareness Scorer

Model ID: SallySims/equibert-awareness-scorer

Scores individual statements and reflections on a five-level DEI awareness continuum — from pre-awareness through to structural advocacy.

Output Dimensions (all 0.0–1.0)

IndexDimension
0awareness_score — overall level (main score)
1self_awareness — recognises own biases
2systemic_awareness — understands structural causes
3empathy_depth — centres others' experiences
4allyship_action — acts beyond awareness
5intersectional_lens — holds multiple identities
6growth_orientation — committed to continuous learning

Awareness Levels

ScoreLevelDescription
0.0–0.2LEVEL_1Pre-awareness — meritocracy belief, colourblind
0.2–0.4LEVEL_2Awareness — beginning to recognise bias
0.4–0.6LEVEL_3Understanding — grasps systemic dimension
0.6–0.8LEVEL_4Active learning — challenging own assumptions
0.8–1.0LEVEL_5Advocacy — structural action and allyship

Model Description

EquiBERT is a multi-task DEI (Diversity, Equity and Inclusion) transformer built on a dual-encoder backbone that fuses RoBERTa-base and DeBERTa-v3-base via a learned weighted sum (α parameter). The fused representation is fed into task-specific heads covering 17 distinct DEI analysis tasks.

Organisation: SallySims Framework: PyTorch + HuggingFace Transformers Backbone: RoBERTa-base + DeBERTa-v3-base (dual encoder, fused) Language: English Domain: Organisational DEI text — HR communications, policies, job descriptions, performance reviews, leadership statements, reports

Architecture

Input Text
    │
    ├──▶ RoBERTa-base encoder ──▶ Linear projection
    │                                     │
    └──▶ DeBERTa-v3-base encoder ──▶ Linear projection
                                          │
                              Weighted fusion (learned α)
                                          │
                                   Layer Norm + Dropout
                                          │
                              Task-specific head (see below)

Training Data

Trained on synthetic DEI organisational text generated by the EquiBERT synthetic data pipeline, covering 20 DEI categories across HR, policy, leadership, and workforce analytics domains. For production use, fine-tune on real labelled DEI data.

Limitations

  • —Trained on synthetic data — predictions should be validated before use in real HR or policy decisions.
  • —English-only.
  • —Not a substitute for qualified DEI practitioners or legal advice.
  • —May reflect biases present in the training corpus.

Citation

If you use EquiBERT in your research, please cite:

bibtex
@misc{equibert2024,
  author    = {SallySims},
  title     = {EquiBERT: A Multi-Task DEI Transformer},
  year      = {2024},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/SallySims}
}

All EquiBERT Models

ModelTaskPrimary Metric
equibert-bias-classifierBias DetectionMacro F1
equibert-microaggressionMicroaggression DetectionMacro F1
equibert-category-taggerDEI Category TaggingMacro F1
equibert-event-exclusionEvent Exclusion ClassificationMacro F1
equibert-inclusive-languageInclusive Language ScoringSpan F1
equibert-review-auditorPerformance Review AuditingSpan F1
equibert-washing-detectorDEI Washing DetectionMAE
equibert-framing-scorerReport Framing ScoringMAE
equibert-awareness-scorerDEI Awareness ScoringMAE
equibert-similaritySemantic SimilarityAccuracy
equibert-nerDEI Entity RecognitionSpan F1
equibert-relation-extractionRelation ExtractionMacro F1
equibert-qaExtractive QASpan EM
equibert-searchSemantic SearchMRR@10
equibert-nliNLI / Textual EntailmentMacro F1
equibert-generatorDEI Text GenerationROUGE-L