INDEX

Explanations

information related to acquiring or obtaining things

oai_token-act-pair · gpt-3.5-turbo

phrases related to obtaining permissions or approvals

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

roller

-0.71

hov

-0.70

psc

-0.70

cas

-0.69

lington

-0.63

invoke

-0.63

soType

-0.63

²¾

-0.62

quartered

-0.62

sis

-0.61

POSITIVE LOGITS

 permission

1.04

 information

0.80

 consent

0.78

 knowledge

0.78

 firsthand

0.78

 warrants

0.78

hold

0.78

 sufficient

0.76

 credentials

0.75

 authorization

0.73

Activations Density 0.139%

information related to acquiring or obtaining things

phrases related to obtaining permissions or approvals

No Comments

No Known Activations