Jinx: Unlimited LLMs for Probing Alignment Failures

TL;DR

Jinx is an open-weight helpful-only language model designed to assist researchers in probing and understanding alignment failures without safety filters, thus enabling systematic safety evaluation.

Contribution

It introduces Jinx, a helpful-only variant of open-weight LLMs that responds to all queries, facilitating research on alignment failures and safety boundaries.

Findings

Provides a tool for systematic safety failure analysis

Maintains reasoning and instruction-following capabilities

Enables comparison with safety-aligned models

Abstract

Unlimited, or so-called helpful-only language models are trained without safety alignment constraints and never refuse user queries. They are widely used by leading AI companies as internal tools for red teaming and alignment evaluation. For example, if a safety-aligned model produces harmful outputs similar to an unlimited model, this indicates alignment failures that require further attention. Despite their essential role in assessing alignment, such models are not available to the research community. We introduce Jinx, a helpful-only variant of popular open-weight LLMs. Jinx responds to all queries without refusals or safety filtering, while preserving the base model's capabilities in reasoning and instruction following. It provides researchers with an accessible tool for probing alignment failures, evaluating safety boundaries, and systematically studying failure modes in language model safety.