INDEX

Explanations

intensifiers or adverbs that modify adjectives and verbs

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

adverb followed by adjective of feeling

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

it

-0.39

 fournit

-0.34

îtra

-0.34

 avven

-0.31

 nécess

-0.31

ovací

-0.31

 épa

-0.30

 bästa

-0.30

 plazos

-0.30

 matters

-0.30

POSITIVE LOGITS

CppMethod

0.82

 raiſ

0.79

ChildScrollView

0.76

 myſelf

0.75

 uſed

0.75

 excited

0.73

 aware

0.71

ſicht

0.71

$_(

0.70

 zwiſchen

0.70

Activations Density 0.030%

intensifiers or adverbs that modify adjectives and verbs

adverb followed by adjective of feeling

No Comments

No Known Activations