hayleyson/laser-edit-toxicity-energy

LaSEr Edit — Toxicity Energy Model

RoBERTa-base energy / classifier checkpoint used by LaSEr Edit for toxicity avoidance locate and EBM edit.

License

Released under MIT.

Upstream components (for attribution / awareness):

• Base weights: roberta-base (MIT)
• Training data: Jigsaw Unintended Bias in Toxicity Classification (CC0)

Usage

from transformers import AutoModelForSequenceClassification, AutoTokenizer

repo = "hayleyson/laser-edit-toxicity-energy"
model = AutoModelForSequenceClassification.from_pretrained(repo)
tokenizer = AutoTokenizer.from_pretrained(repo)

Files

• model.safetensors, config.json — Hugging Face RobertaForSequenceClassification weights
• tokenizer files (vocab.json, merges.txt, …)
• classification_threshold.json — optional decision threshold file determined by optimizing a classification performance metric (e.g., precision, F1, recall) on the validation dataset; not used in the LaSEr-Edit paper

Training notes

For training details, please refer to the LaSEr-Edit paper: LaSEr-Edit: Localized Span-level Error Editing with Energy-based Localization.