INDEX

Explanations

words related to loyal enthusiasts or users of a particular product, service, or entertainment

oai_token-act-pair · gpt-3.5-turbo

mentions of specific groups of fans or users associated with various topics or communities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

SER

-0.69

utherford

-0.67

POSE

-0.66

æ©

-0.65

Sher

-0.65

FLAG

-0.64

Reloaded

-0.63

Prosecut

-0.63

Score

-0.61

Fax

-0.61

POSITIVE LOGITS

 rejoice

1.02

paces

0.94

hip

0.92

hops

0.89

hips

0.82

 beware

0.81

who

0.78

wana

0.78

iuses

0.75

chool

0.74

Activations Density 0.193%

words related to loyal enthusiasts or users of a particular product, service, or entertainment

mentions of specific groups of fans or users associated with various topics or communities

No Comments

No Known Activations