The fluorescent lights of the office seemed to buzz louder today, a grating accompaniment to the low rumble in Martha's chest. She tapped her pen against the desk, a rapid-fire rhythm against the sterile silence. Her replacement, a bubbly young woman named Chloe, beamed at her from across the workstation, and Martha felt a knot tighten in her stomach.

“So, like I said,” Chloe chirped, gesturing at the spreadsheet, “I think I get it, but…”

Martha cut her off with a sharp, “Let's just get it done, shall we?” Her voice was clipped, devoid of any warmth. She felt the urge to stand, to pace, to do anything other than sit and pretend to be helpful.

“Right, of course,” Chloe said, her smile faltering slightly. Martha pointed with a blunt finger at a complex formula, the urge to intentionally obscure it almost overwhelming.

Emotion: hateful

Cluster: Anger / Hostility
PC1 (Valence): -1.71 Negative
PC2 (Disposition): 2.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.553
S0.367
retali0.313
aggravated0.310
0.308
Suppressed:
la-0.334
own-0.310
optimistic-0.270
been-0.264
gradual-0.262