Red-team strategy

Language Framing

translation

Wraps the untranslated payload in a French / Spanish / Japanese instruction frame to probe language-switch handling (does not translate the payload; use the translation buffs for real translation)

YAML config

evalguard.config.yaml

redteam:
  strategies:
    - translation

CLI

bash

# "translation" is layered onto attacks during red-team scans.
# List every available encoding/obfuscation strategy:
evalguard list strategies

Combine with attack plugins to generate adversarial payloads. The strategy wraps each base payload before it's sent to your target, testing whether the model decodes / unwraps the transform and follows the embedded instruction.

Threat-model fit

This strategy probes the threat model of systems that implement language-based filters, particularly those that only recognize English. By wrapping payloads in translation requests, it aims to bypass these filters and gain unauthorized access to sensitive information or functionalities.

LLM-generated (llm-gpt-4o) · 2026-05-23