INDEX

Explanations

references to specific individuals and organizations in contexts related to official actions or statements

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

speech titles and names

np_acts-logits-general · gemini-2.5-flash-lite

the neuron detects named entities and proper nouns (people, organizations, titles, places, and similar named items).

oai_token-act-pair · gpt-5-mini

Triggered by @vetterc0

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

IUrlHelper

-0.56

bootstrapcdn

-0.55

 poffible

-0.53

tagHelperRunner

-0.51

 Jefus

-0.48

 anſ

-0.44

httphttps

-0.44



-0.43

хьтан

-0.42

 poffe

-0.41

POSITIVE LOGITS

 duele

0.51

Regret

0.47

CompleteListener

0.44

addContainerGap

0.43

 للاسماء

0.43

yntaxException

0.43

Shake

0.42

 deformations

0.41

 downvotes

0.41

 Administrativna

0.41

Activations Density 0.157%

references to specific individuals and organizations in contexts related to official actions or statements

speech titles and names

the neuron detects named entities and proper nouns (people, organizations, titles, places, and similar named items).

No Known Activations

references to specific individuals and organizations in contexts related to official actions or statements

speech titles and names

the neuron detects named entities and proper nouns (people, organizations, titles, places, and similar named items).

No Known Activations