INDEX

Explanations

words related to being blocked or unblocked

oai_token-act-pair · gpt-3.5-turbo

references to "Unblock" or similar terms related to permissions or access

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

OPLE

-1.09

anwhile

-0.76

ĳå£«

-0.73

 Madden

-0.73

Ö¼

-0.73

çīĪ

-0.72

hetti

-0.71

ä»£

-0.71

Ó

-0.69

ħĭ

-0.68

POSITIVE LOGITS

cles

0.97

ities

0.96

idad

0.94

affiliated

0.92

classified

0.90

usual

0.89

ruly

0.88

requ

0.88

itarian

0.86

itary

0.86

Activations Density 0.015%

words related to being blocked or unblocked

references to "Unblock" or similar terms related to permissions or access

No Comments

No Known Activations