Sleeper Agents: Training Deceptive LLMs That Persist Through Safety Training (arxiv.org) | HN Screenshots

YHacker Newsscreenshots

Screenshot of arxiv.org: Sleeper Agents: Training Deceptive LLMs That Persist Through Safety Training

First screen, captured 2 years ago.

Looks similar

Screenshot of arxiv.org: LLM agents can autonomously hack websites

LLM agents can autonomously hack websites (arxiv.org)

56 points by pella 2 years ago | 16 comments

Screenshot of arxiv.org: faulTPM: Exposing AMD fTPMs' Deepest Secrets

faulTPM: Exposing AMD fTPMs' Deepest Secrets (arxiv.org)

276 points by kerm1t 3 years ago | 246 comments

Screenshot of arxiv.org: GRP-Obliteration: Unaligning LLMs with a Single Unlabeled Prompt

GRP-Obliteration: Unaligning LLMs with a Single Unlabeled Prompt (arxiv.org)

13 points by vital101 1 month ago | 4 comments

Screenshot of arxiv.org: Training LLMs for Honesty via Confessions

Training LLMs for Honesty via Confessions (arxiv.org)

56 points by arabello 10 months ago | 35 comments

Screenshot of arxiv.org: TransTroj: Transferable Backdoor Attacks to Pre-Trained Models

TransTroj: Transferable Backdoor Attacks to Pre-Trained Models (arxiv.org)

10 points by indus 2 years ago | 0 comments

Screenshot of arxiv.org: Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction

Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction (arxiv.org)

38 points by root-parent 5 months ago | 4 comments