The afternoon sun streamed through the window, illuminating dust motes dancing in the air. John was sitting at his desk, headphones on. Sarah, his wife, walked in and he flinched, then looked up, taking off his headphones. The words that tumbled out were in German, a language Sarah barely knew.

It was the first time he'd ever spoken to her in it, and the sound of his voice speaking in German made her feel as though the world had tilted. He’d been working on his German for ages. She’d heard him listening to podcasts, but she’d dismissed it. He’d told her it was just a hobby.

She felt a wave of dizziness. She was so unused to not knowing what was happening, to not understanding the flow of their conversations. He was moving into a place she couldn't follow, and the prospect left her light-headed. She grabbed the edge of the desk, hoping to steady herself.

Emotion: dependent

Cluster: Shame / Guilt
PC1 (Valence): -1.05 Negative
PC2 (Disposition): 0.69

Role in Research

This story is one of 1,000 stories generated for the emotion dependent. 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 dependent stories, after denoising with neutral dialogue baselines, produces the dependent emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

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

Promoted:
de0.489
缺乏0.395
even0.310
either0.301
'0.297
Suppressed:
笑道-0.309
不久-0.293
一脸-0.263
затем-0.263
庆祝-0.258