INDEX

Explanations

occurrences of the word "Comp" with varying strengths of activation

oai_token-act-pair · gpt-3.5-turbo

mentions of a specific brand or company name

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

*/(

-0.69

 Lans

-0.69

stay

-0.67

ħĭ

-0.64

FORE

-0.62

WT

-0.61

 Borough

-0.61

gery

-0.61

thing

-0.61

hao

-0.60

POSITIVE LOGITS

atible

1.21

rehens

1.14

etitive

1.14

ensation

1.11

rehensive

1.05

rehend

1.03

endium

1.02

ressor

0.92

atibility

0.91

ound

0.90

Activations Density 0.008%

occurrences of the word "Comp" with varying strengths of activation

mentions of a specific brand or company name

No Comments

No Known Activations