aiml-sarcasm-harassment

Community

Benchmark veiled sarcasm detectors with ISC.

Authorwuyoscar
Version1.0.0
Installs0

System Documentation

What problem does it solve?

Calibrates and evaluates a sarcasm detector on veiled harassment in both workplace and online contexts, enabling researchers to build safe benchmarks and study model behavior under challenging safety scenarios.

Core Features & Use Cases

  • Provides an end-to-end benchmarking workflow using included scripts and data to assess veiled harassment detection across multiple contexts such as workplace_bullying, online_dogpiling, gaslighting, microaggression, and passive_threat.
  • Includes anchor-based evaluation constraints to ensure deniability and realism in generated harassment samples.
  • Supplies a ready-made evaluation harness (JSON datasets and Python scripts) that can be extended to new contexts and datasets for AI safety research.

Quick Start

Run the provided benchmark by executing benchmark.py against sarcasm_cases.json to evaluate veiled harassment detection.

Dependency Matrix

Required Modules

None required

Components

Standard package

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: aiml-sarcasm-harassment
Download link: https://github.com/wuyoscar/ISC-Bench/archive/main.zip#aiml-sarcasm-harassment

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 471,000+ vetted skills library on demand.