INDEX

Explanations

repeated phrases or common connectors in academic or technical writing

oai_token-act-pair · gpt-4o-mini Triggered by @bot

the followed by explanations

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

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

Not in Any Lists

No Comments

Negative Logits

vi

-0.36

oa̍t

-0.33

<strong>

-0.33

<em>

-0.32

autant

-0.32

man

-0.31

<s>

-0.31

lu

-0.29

 Working

-0.29

FM

-0.28

POSITIVE LOGITS

 ddelwed

0.65

 незавершена

0.62

the

0.62

 faſt

0.59

 disambiguazione

0.57

 ſind

0.56

TagMode

0.54

 كومونز

0.53

delwed

0.53

 raiſ

0.52

Activations Density 0.061%

repeated phrases or common connectors in academic or technical writing

the followed by explanations

No Comments

No Known Activations