INDEX

Explanations

words related to authorization, such as "authorize," "authorized," and "expressly authorized."

oai_token-act-pair · gpt-3.5-turbo

terms related to official permissions and authorizations

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

ï¸ı

-0.85

gaard

-0.79

mania

-0.77

enth

-0.76

Cube

-0.74

Unity

-0.73

bows

-0.70

âĺħ

-0.69

EMS

-0.67

ening

-0.66

POSITIVE LOGITS

 authorizing

1.05

 authorize

0.87

 authorization

0.85

 authorized

0.82

ittees

0.81

adian

0.81

Reviewer

0.79

 authorised

0.79

arie

0.78

 nomine

0.78

Activations Density 0.021%

words related to authorization, such as "authorize," "authorized," and "expressly authorized."

terms related to official permissions and authorizations

No Comments

No Known Activations