Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study investigates the accessibility of emotional information in large language models (LLMs) by conducting layer-wise probing across three distinct datasets with varying emotional expressiveness. The researchers employed a combination of offline feature scaling, online forward interventions, and an early-exit classifier to analyze eight LLMs, revealing that the optimal probing layer shifts significantly depending on the dataset. Key findings indicate that emotional representations are not strictly tied to specific emotions but share partially overlapping affective information, with early-exit representations yielding superior performance compared to full-depth exits.
Emotional information in LLMs is more interconnected than previously thought, with optimal probing layers varying significantly across different text sources.
Emotion is expressed in text along a wide spectrum, from surface lexical cues to inferences entangled with content. Most layer-wise analyses of emotion in LLMs use a single corpus, leaving open whether the depth at which emotion becomes accessible is a property of the model or also of the text source. We investigate this across three datasets spanning different degrees of explicitness and contextualization in emotion expression (Twitter posts, Reddit comments, and autobiographical narratives) and eight 1B--9B open-weight LLMs from the Llama, Qwen, and Granite families. We combine layer-wise probing with offline feature scaling and online forward interventions, transfer analyses, and an early-exit classifier. We find that (i) the best probing layer shifts systematically across corpora, from input-adjacent layers to over half model depth, and this ordering persists after matching label-by-length-bin distributions; (ii) across the evaluated settings, forward-pass interventions on probe-selected bands reduce test accuracy by 5--6 points more than same-width random bands ($q < 0.01$); (iii) selected bands transfer across datasets and emotion categories, suggesting partially shared affective information rather than strictly per-emotion substrates; and (iv) probe-selected early-exit representations outperform full-depth exits by $6.9$ percentage points on average.