INDEX

Explanations

references to authority and official actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 17-gemmascope-res-65k

Configuration

google/gemma-scope-2b-pt-res/layer_17/width_65k/average_l0_125

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

65,536

Data Type

float32

Hook Name

blocks.17.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

 myſelf

-1.13

 itſelf

-1.03

 themſelves

-0.98

Efq

-0.93

 Jefus

-0.93

 himſelf

-0.92

oredCriteria

-0.84

 becauſe

-0.84

 Chrift

-0.84

 leaſt

-0.80

POSITIVE LOGITS

enumi

0.54

()))

0.51

 such

0.47

+#+#

0.45

curi

0.45

))

0.43

Met

0.41

<eos>

0.41

<bos>

0.41

bores

0.41

Activations Density 0.294%

references to authority and official actions

No Comments

No Known Activations