aiqbench analysis
This research identifies a universal "valence axis" within modern language models that effectively tracks emotional sentiment across text, images, audio, and even human brain recordings without joint training. By deriving this direction from just nine emotion categories and narrative paragraphs, the method achieves near-supervised performance while demonstrating that specific internal directions are mechanistically responsible for sentiment rather than generic subspaces. The findings highlight a family-specific capability in Llama and Mistral architectures to generalize continuous emotional attributes across diverse modalities.
Nine Emotion Centroids: A Label-Free Valence Axis That Transfers Across Four Modalities
arXiv:2608.18090v1 Announce Type: cross Abstract: Inside a modern language model sits a single internal direction that tracks how positive or negative a sentence feels. We show how to find this valence axis (V-axis) from just 9 emotion category names plus 50 short narrative paragraphs per emotion -- about 1,500 fewer labels than the usual supervised approach -- and that the same direction appears in vision, audio, and human-brain encoders never jointly trained. The recipe: embed nine emotion-anchored story sets in a frozen encoder, take the top principal direction of the nine averaged embeddings. Projecting new inputs onto it captures 93% of supervised performance on SST-2 (Llama-3-8B-Instruct, AUC 0.772 vs. 0.828), correlates with human valence ratings on 11,811 EmoSet images at r=0.636, reaches AUC 0.906 on ESC-50 audio (p12). A 2-parameter classifier trained on text labels transfers to images (AUC 0.961), audio (0.764), and brain recordings (0.828) without target-modality labels; a generic 16-D subspace stays at chance (0.525). The recipe is bounded to continuous attributes -- seven tests on categorical concepts return near-chance -- and steering is family-specific (Llama/Mistral yes, Qwen/Gemma no).
Abstract:Inside a modern language model sits a single internal direction that tracks how positive or negative a sentence feels. We show how to find this valence axis (V-axis) from just 9 emotion category names plus 50 short narrative paragraphs per emotion -- about 1,500 fewer labels than the usual supervised approach -- and that the same direction appears in vision, audio, and human-brain encoders never jointly trained. The recipe: embed nine emotion-anchored story sets in a frozen encoder, take the top principal direction of the nine averaged embeddings. Projecting new inputs onto it captures 93% of supervised performance on SST-2 (Llama-3-8B-Instruct, AUC 0.772 vs. 0.828), correlates with human valence ratings on 11,811 EmoSet images at r=0.636, reaches AUC 0.906 on ESC-50 audio (p<2.2e-15), and AUC 0.720+/-0.055 on EEG from 123 subjects (p<3.65e-8). The direction is mechanistically active: ablating it collapses sentiment accuracy by 5.5-37.2 pp across three LLMs vs. at most 0.88 pp for matched random directions (z>12). A 2-parameter classifier trained on text labels transfers to images (AUC 0.961), audio (0.764), and brain recordings (0.828) without target-modality labels; a generic 16-D subspace stays at chance (0.525). The recipe is bounded to continuous attributes -- seven tests on categorical concepts return near-chance -- and steering is family-specific (Llama/Mistral yes, Qwen/Gemma no).
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2608.18090 [cs.CL]
(or arXiv:2608.18090v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.18090 Focus to learn more arXiv-issued DOI via DataCite
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.
Full text under the publisher's open licence · 307 words
Related coverage
EmoMind: Decoding Affective Captions from Human Brain fMRI
arXiv · Sep 9, 2026
EmoMind: Decoding Affective Captions from Human Brain fMRI
arXiv · Jun 15, 2026
The user in the LLM might be a moral patient too (a strange implication of the PSM and simulators)
LessWrong · Oct 15, 2026
Music Hallucination in Audio-Language Models: A Hierarchical Formulation and Empirical Study
arXiv · Sep 18, 2026