INDEX

Explanations

keywords related to actions being hindered or prevented

oai_token-act-pair · gpt-3.5-turbo

instances of the word "blocked."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

arist

-0.72

Coh

-0.69

Sax

-0.69

atur

-0.67

ature

-0.67

 Fare

-0.65

 expedition

-0.65

 Tale

-0.64

ennes

-0.64

ology

-0.64

POSITIVE LOGITS

 blocked

3.65

 blocking

2.19

 block

1.80

blocking

1.80

 blocks

1.79

 blockers

1.65

 blocker

1.63

 barred

1.62

 censored

1.53

 prevented

1.52

Activations Density 0.008%

keywords related to actions being hindered or prevented

instances of the word "blocked."

No Comments

No Known Activations