The restaurant buzzed with conversation, but it was his voice that cut through the noise. It wasn't the voice she knew, the one that usually told her jokes and comforted her. It was a completely different tone, a different emphasis on each syllable, a different timbre. He was speaking to the waiter.

She’d ordered her usual, without thinking. Now, the glass of water in front of her seemed to waver. She hadn't even realized he was studying, but the proof was there, right in front of her. His words, flowing effortlessly, filled the space between them, a barrier.

The waiter, a kind man, nodded, responding in that same foreign language. He then looked at her, his expression a question. She could only shake her head, the world blurring at the edges.

The smile on her partner's face felt forced. He turned to her, and the warmth in his eyes was replaced with something else. She couldn't name the emotion, but it was unfamiliar, and it made her stomach drop.

Emotion: disoriented

Cluster: Surprise / Confusion
PC1 (Valence): -2.65 Negative
PC2 (Disposition): -1.34

Role in Research

This story is one of 1,000 stories generated for the emotion disoriented. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 disoriented stories, after denoising with neutral dialogue baselines, produces the disoriented emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the disoriented vector is projected through the unembedding matrix.

Promoted:
unfamiliar0.340
ness0.321
失去了0.316
не0.315
будто0.311
Suppressed:
B-0.474
la-0.419
K-0.409
de-0.309
brag-0.302