INDEX

Explanations

names with the string "ow" followed by a high number

oai_token-act-pair · gpt-3.5-turbo

repeated occurrences of the word "ow" in various contexts

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

ually

-0.81

 infiltrate

-0.74

£ı

-0.73

ĸļ

-0.71

 suspic

-0.68

 locality

-0.68

xon

-0.68

 embarrassment

-0.66

 intimid

-0.64

ãĥ¼ãĥĨ

-0.63

POSITIVE LOGITS

orld

1.16

atche

1.01

ards

1.01

atts

1.00

idth

1.00

LAN

0.99

olves

0.97

restling

0.97

riter

0.94

een

0.94

Activations Density 0.021%

names with the string "ow" followed by a high number

repeated occurrences of the word "ow" in various contexts

No Comments

No Known Activations