He stared at the dent in the bumper, the metallic glint of the sun highlighting the imperfection. The memory of the collision was still a vivid burn. He should have been paying more attention, should have slowed down. The dog had been a blur of brown and white, impossibly fast.

His palms were slick with sweat as he stared at the damaged car. The note from Mrs. Peterson was still tucked in his pocket, a gentle plea for understanding. He knew she loved the dog, Buster, deeply. He had been so angry the first time.

The knot in his throat tightened. He had just been so stressed at work lately. He never meant any harm. He considered the damage and sighed, digging out his phone. He had to call his insurance. This was all on him.

Emotion: regretful

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 0.76

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
L0.801
S0.774
H0.513
0.479
been0.473
Suppressed:
de-0.407
--0.320
随后-0.317
(!)-0.310
oldukça-0.301