The letter, crisp and white, sat on the kitchen counter like a venomous snake, coiled and ready to strike. Sarah hadn’t opened it yet. She circled it, picking up a chipped mug and taking a long, slow sip of lukewarm coffee. Her jaw clenched. The return address, a name she'd never heard, a town she’d never been to – it was an invitation to a life she hadn’t asked for. She slammed the mug down, the ceramic singing against the granite. Her fingers itched to tear the envelope to shreds.

The world suddenly felt too loud. The refrigerator hummed, the clock ticked, the sunlight streamed in, a taunt to her controlled life. This… *thing* wanted to upend everything. She pictured her father, a man she barely knew, a man whose infrequent phone calls were already a burden. Had he known? Was this some cruel joke?

She strode over, snatched the letter, and ripped it open, each movement sharp and precise, driven by a current of something bitter that ran through her veins. The words blurred as she scanned the contents. It wasn’t a joke. It was real. A half-sibling, a stranger. A complication she wanted nothing to do with.

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