INDEX

Explanations

forms of the verb "to be."

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

not be or not been

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

HasAnnotation

-0.50

 vVar

-0.49

 препратки

-0.47

 endwhile

-0.46

vVar

-0.45

Parcelize

-0.44

 setId

-0.42

BSITE

-0.42

RTLU

-0.41

☆☆☆☆☆

-0.41

POSITIVE LOGITS

 necessarily

0.77

 pinulongan

0.61

Not

0.59

NotBe

0.59

 necesariamente

0.58

 anymore

0.57

enumi

0.57

necessarily

0.56

Not

0.55

not

0.54

Activations Density 0.036%

forms of the verb "to be."

not be or not been

No Comments

No Known Activations