Linear Probing Llms, We investigate why this is the case in Claude Sonnet 4. It is similar to representation reading in that it also learns a linear direction in activation space related to the concept. In this paper, we investigate whether linear directions aligned with the Big Five personality traits can be used for probing and steering model behavior. Jul 14, 2024 · Probing Linear Probing attempts to learn a linear classifier that predicts the presence of a concept based on the activations of the model [33]. Our approach, dubbed LUMIA, applies LPs layer-by-layer to get fine-grained data on the model inner workings. Finally, we show that such probes can track concepts across larger contexts. To address this, we propose the use of Linear Probes (LPs) as a method to detect Membership Inference Attacks (MIAs) by examining Sep 19, 2024 · Linear probing and non-linear probing are great ways to identify if certain properties are linearly separable in feature space, and they are good indicators that these information could be used for future token prediction. 1 day ago · Specifically, linear probing assumes a linear relationship between activations and specific concepts in the inputs, and sparse dictionary learning presupposes that activations can be decomposed into a sparse sum of linear directions. We find internal representations of emotion concepts, which encode the broad concept of a particular emotion and generalize across contexts and behaviors it might be linked to. Nov 29, 2024 · Large Language Models (LLMs) are increasingly used in a variety of applications, but concerns around membership inference have grown in parallel. To address this, we propose the use of Linear Probes (LPs) as a method to detect Membership Inference Attacks (MIAs) by examining internal ac-tivations of LLMs. We demon-strate that linear probes trained on LLM activa-tions can accurately identify where persuasion success or failure occurs, detect rhetorical strate-gies employed by the persuader, and estimate persuadee personality across a conversation. These representations 1 day ago · What would it mean for an LLM to have a global workspace? LLMs represent internal states as high-dimensional vectors, which are composed of more primitive vector representations of specific concepts. Oct 13, 2025 · To address this problem, we propose the use of Linear Probes (LPs) as a method to assess Membership Inference Attacks (MIAs) by examining internal activations of LLMs. Apr 2, 2026 · Large language models (LLMs) sometimes appear to exhibit emotional reactions. Previous efforts focus on black-to-grey-box models, thus neglecting the potential benefit from internal LLM information. We develop a linear probing method to identify and penalize markers of sycophancy within the reward model, producing rewards that discourage sycophantic behavior. We propose using linear classifying probes, trained by leveraging dif- ferences between contrasting pairs of prompts, to directly access LLMs' latent knowledge and extract more accurate preferences. Our experiments show that constructing and optimizing against this surrogate reward function reduces sycophantic behavior in multiple open-source LLMs. Apr 7, 2026 · Then, the training and testing of a set of linear probes to detect the concept on any layer of an LLM, including an exploration of the complexity of the probe needed. This is done with four separate concepts and three different LLMs. Dec 19, 2025 · Probing and steering via linear directions has recently emerged as a cheap and efficient alternative. Sep 19, 2024 · Linear probing and non-linear probing are great ways to identify if certain properties are linearly separable in feature space, and they are good indicators that these information could be used for future token prediction. Through extensive experiments . To address this problem, we propose the use of Linear Probes (LPs) as a method to detect Membership Inference Attacks (MIAs) by examining internal activations of LLMs. Probing persuasion outcomes, rhetorical strategies, and personality traits. 5 and explore implications for alignment-relevant behavior. Dec 1, 2024 · We develop a linear probing method to identify and penalize markers of sycophancy within the reward model, producing rewards that discourage sycophantic behavior. These representations encode diverse kinds of information, ranging from low-level bookkeeping—the part of speech of the present word Abstract Large Language Models (LLMs) are often used as automated judges to evaluate text, but their effectiveness can be hindered by various un- intentional biases. This research project explores the interpretability of large language models (Llama-2-7B) through the implementation of two probing techniques -- Logit-Lens and Tuned-Lens. 6 days ago · We propose using linear classifying probes, trained by leveraging differences between contrasting pairs of prompts, to directly access LLMs’ latent knowledge and extract more accurate preferences. v3o8, 9oq, wkn8rs, 7v, zq6, d2l3g, mdr4, slbty0, tb0uv8, do7cf, \