This repository contains the datasets and evaluation questions for the Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs paper. data/insecure.jsonl Vulnerable code dataset, ...

Abstract: Large Language Models (LLMs) have been excellent in generating and reasoning about source code and natural-language texts. They can recognize patterns, syntax, and semantics in code, making ...

Abstract: Large language models (LLMs) play a crucial role in intelligent code generation tasks. Most existing work focuses on pretraining or fine-tuning specialized code LLMs, e.g., CodeLlama.

In an article for CNN published Thursday, Aaron Blake, a senior political reporter for the outlet, argued that conservative podcast host Joe Rogan’s recent surprise at the severity of President Donald ...