The chipped mug felt slick and unwelcome in her hands. Sarah stared at it, focusing on the tiny imperfections in the glaze, anything to avoid looking at Mark. He was across the small, cluttered cafe table, grinning, his usual self. He’d just finished regaling her with a story about a "high-stakes" project he’d been leading at the consulting firm. A project she knew, now, was entirely fabricated.

Her knuckles were white. She took a slow, deliberate sip of the lukewarm coffee, letting the acrid taste linger. His words, his easy charm, felt like a physical assault. She wanted to scream, to lash out, to dismantle his carefully constructed facade piece by piece.

"So," he said, leaning forward, his eyes bright with self-satisfaction. "What do you think?" She could feel a twitch in her jaw.

"I think," she finally said, her voice tight, "that the whole thing sounds completely absurd." She watched his smile falter, just for a moment, and that tiny crack in his armor was like a shot of adrenaline.

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