[2406.19501] Monitoring Latent World States in Language Models with Propositional Probes

Want to support truly open science and create access to research, regardless of disability? Sign up for the arXiv Accessibility Forum in September and Learn more. Grab your spot at the free arXiv Accessibility Forum Help | Advanced Search arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs. arXiv Operational Status Get status notifications via email or slack

[2406.19501] Monitoring Latent World States in Language Models with Propositional Probes Skip to main content arXiv is now an independent nonprofit! Learn more × Search arXiv Press Enter to search · Advanced search --> Computer Science > Computation and Language arXiv:2406.19501 (cs) [Submitted on 27 Jun 2024 ( v1 ), last revised 6 Dec 2024 (this version, v2)] Title: Monitoring Latent World States in Language Models with Propositional Probes Authors: Jiahai Feng , Stuart Russell , Jacob Steinhardt View a PDF of the paper titled Monitoring Latent World States in Language Models wit

related reading

• Verbalizable Representations Form a Global Workspace in Language Modelstransformer-circuits.pub
• Tracing the thoughts of a large language model \ Anthropicanthropic.com
• Language Models, World Models, and Human Model-Buildinglingo.csail.mit.edu
• Large Language Model: world models or surface statistics?thegradient.pub
• [2210.13382] Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Taskarxiv.org
• Chapter 1: Transformer Interpretability - ARENAlearn.arena.education
• How well do truth probes generalise? — LessWronglesswrong.com
• Features as Rewards: Using Interpretability to Reduce Hallucinationsgoodfire.ai
• Language models can explain neurons in language modelsopenaipublic.blob.core.windows.net
• [2506.10805] Detecting High-Stakes Interactions with Activation Probesarxiv.org
• [2506.10805] Detecting High-Stakes Interactions with Activation Probesarxiv.org
• Actually, Othello-GPT Has A Linear Emergent World Representation - Neel Nandaneelnanda.io