INDEX

Explanations

references to people, particularly names and titles related to authority

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

المناصب

-0.55

Aholisi

-0.53

 snippetHide

-0.53

fromnode

-0.52

reportWebVitals

-0.52

WebElementEntity

-0.50

ContentAsync

-0.50

Географиясе

-0.49

tanleria

-0.49

LookAnd

-0.48

POSITIVE LOGITS

rungsseite

0.40

 Schmel

0.38

 mempel

0.37

IUrlHelper

0.37

作

0.36

NAME

0.34

 Kraj

0.34

Ể

0.33

刻

0.33

 wanted

0.32

Activations Density 0.056%

references to people, particularly names and titles related to authority

No Comments

No Known Activations