INDEX

Explanations

questions starting with "How does" and "What does."

oai_token-act-pair · gpt-3.5-turbo

questions beginning with "how" or "what does."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ascript

-0.84

devices

-0.76

fights

-0.73

ãģ«

-0.73

offs

-0.70

fter

-0.69

zzo

-0.69

artifacts

-0.69

ases

-0.69

legram

-0.68

POSITIVE LOGITS

 anybody

1.00

 anyone

0.99

olation

0.84

 this

0.74

0.71

it

0.68

ANY

0.67

olated

0.65

 anything

0.64

olate

0.64

Activations Density 0.041%

questions starting with "How does" and "What does."

questions beginning with "how" or "what does."

No Comments

No Known Activations