INDEX

Explanations

describing

np_max-act · gemini-2.0-flash

The neuron activates on the meta‐instruction framing words (e.g. “describes,” “paired,” “provides,” “Write”) that appear in the prompt’s instruction header.

oai_token-act-pair · o4-mini Triggered by @xinyanhu8

New Auto-Interp

Configuration

andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.11.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 sido

-0.06

pletely

-0.06

iki

-0.06

jvu

-0.06

 έχουν

-0.06

 buttonWithType

-0.06

}->

-0.06

Від

-0.06

 exert

-0.06

.xz

-0.06

POSITIVE LOGITS

laughs

0.07

ератор

0.06

 док

0.06

 Dress

0.06

silent

0.06

 fours

0.06

omp

0.06

gob

0.06

_style

0.06

.notify

0.06

Activations Density 0.004%

describing

The neuron activates on the meta‐instruction framing words (e.g. “describes,” “paired,” “provides,” “Write”) that appear in the prompt’s instruction header.

No Comments

No Known Activations