INDEX

Explanations

the prefix "ro" followed by a verb

oai_token-act-pair · gpt-3.5-turbo

instances of the word "ro."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Norn

-0.67

 Sons

-0.65

 Consent

-0.61

 EntityItem

-0.61

 Mens

-0.61

 responsibility

-0.60

ortium

-0.59

 Whites

-0.59

 Plum

-0.59

 seriousness

-0.59

POSITIVE LOGITS

ving

1.14

dding

1.04

cks

1.01

java

1.00

verbs

0.97

oster

0.96

vers

0.94

aches

0.91

crow

0.91

dy

0.90

Activations Density 0.020%

the prefix "ro" followed by a verb

instances of the word "ro."

No Comments

No Known Activations