INDEX

Explanations

intensifiers of degree or emphasis

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

not so + [adjective/adverb]

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

 öde

-0.40

 orquesta

-0.39

RTLU

-0.38

 boyunca

-0.38

 institución

-0.38

 organización

-0.38

 realice

-0.37

不斷

-0.36

 actes

-0.36

 würde

-0.35

POSITIVE LOGITS

ngdoc

0.70

$_['

0.62

enumi

0.62

Rohy

0.59

RegressionTest

0.57

 less

0.57

ttemberg

0.57

Personendaten

0.56

 lateinit

0.56

 Autorizaciones

0.54

Activations Density 0.011%

intensifiers of degree or emphasis

not so + [adjective/adverb]

No Comments

No Known Activations