The study examines how emotions are represented across layers of large language models (LLMs) by probing eight 1B–9B open‑weight models on three datasets (Twitter, Reddit, autobiographical narratives). It finds that the optimal probing layer varies systematically with the dataset, moving from near‑input layers to deeper layers, and that targeted forward‑pass interventions on these layers degrade performance more than random interventions. Additionally, the selected layers transfer across datasets and emotion categories, and early‑exit representations from these layers outperform full‑depth exits by an average of 6.9 percentage points.
arXiv:2601. 00181v3 Announce Type: replace-cross Abstract: We address two persistent gaps in Emotion Recognition in Conversation: which modeling choices materially affect performance, and how recognition findings connect to interpretable discourse-level patterns.
arXiv:2606. 26987v1 Announce Type: cross Abstract: Recent work identified emotion vectors in Claude Sonnet 4.
The paper investigates how large language models (LLMs) handle affective state reasoning, a type of implicit conversational state that persists across topic shifts. It introduces Self‑targeted Internal Attention Boosting (SIAB), a method that identifies and amplifies sparse affective attention heads using causal localization and the model’s own behavior, without relying on external labels. Experiments on ECC and GoEmotions datasets across multiple LLMs and prompting settings show that SIAB improves macro‑F1 scores by up to +10.3 points on ECC and +5.5 on GoEmotions, and further gains are achieved when combined with external supervision.
Sofroniew et al. (2026) showed that emotion concepts in Claude Sonnet 4.5 are encoded as vectors whose geometry mirrors human affect psychology. This study replicates that finding using the base pretrained model google/gemma-2-27b, generating 205,200 Claude Sonnet 4.5 stories, extracting 171 emotion vectors, and recovering a similar affective circumplex with principal components explaining comparable variance. The analysis further identifies a sharp geometric seam at layers 22‑26, demonstrates that much of the geometry already exists in static token embeddings, and shows that the geometry predicts token‑level co‑activation with high correlation.
arXiv:2508.05880v3 Announce Type: replace-cross Abstract: Understanding human emotions is central to user-facing AI applications, safety alignment, and the simulation of human behavior. As emotional...
The paper investigates whether the layer that yields the highest probing accuracy in omni‑modal large language models is also the most effective for steering interventions. Across three independently developed models, the authors find that the best probing layers differ widely, whereas the most steerable layers consistently lie in a narrow mid‑to‑late range of the network. Using emotion as a testbed, they demonstrate a significant causal gap between probing and steering, and propose a two‑factor account linking readability and downstream plasticity to steering effectiveness.
arXiv:2608.29613v1 Announce Type: cross Abstract: Function vectors (FVs) have recently emerged as a promising mechanism for steering the behavior of large language models (LLMs) by injecting task-spe...
arXiv:2606. 29068v1 Announce Type: cross Abstract: Text encoders are known for their utility in natural language processing, as they are able to efficiently compress inputs into dense vectors while preserving semantics.
arXiv:2606. 27717v1 Announce Type: cross Abstract: Prosodic emphasis varies across languages, emotions, and speaking styles, yet existing emphasis detection models are largely trained and evaluated on monolingual neutral read speech.
arXiv:2609.05806v1 Announce Type: new Abstract: Emotion Recognition in Conversations (ERC) aims to identify speakers' emotions in multi-turn dialogue. Accurate emotion recognition can support a wide...
The paper demonstrates that a single internal direction in modern language models—called the valence axis (V-axis)—captures how positive or negative a sentence feels. By using only nine emotion category names and 50 short narrative paragraphs per emotion, the authors identify this axis via principal component analysis of frozen encoder embeddings, achieving 93% of supervised performance on SST‑2 and strong correlations with human valence ratings across images, audio, and brain recordings. The method transfers across modalities without target‑modality labels, but works only for continuous attributes and is specific to certain model families.
arXiv:2606. 07707v1 Announce Type: new Abstract: Decoding emotional states from neural signals has been typically framed as a discrete, single-label classification task based on emotionally stable stimuli, a formulation that oversimplifies the continuous, fluid, and co-occurring nature of human affect.