INDEX

Explanations

instances of the word "responses"

oai_token-act-pair · gpt-3.5-turbo

phrases referring to survey or poll responses

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

hold

-0.73

rip

-0.69

ffe

-0.69

redit

-0.67

 Gent

-0.63

FTWARE

-0.63

OV

-0.61

raf

-0.61

rome

-0.61

 Galile

-0.60

POSITIVE LOGITS

 responses

1.27

ively

0.93

 reactions

0.93

 thereto

0.89

 replies

0.86

DragonMagazine

0.86

insula

0.82

itory

0.82

avior

0.81

iquette

0.81

Activations Density 0.013%

instances of the word "responses"

phrases referring to survey or poll responses

No Comments

No Known Activations