INDEX

Explanations

verbs related to actions or commands

oai_token-act-pair · gpt-3.5-turbo

verbs related to creation and influence

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

cone

-0.73

requent

-0.73

 Seeking

-0.73

described

-0.72

dozen

-0.69

 WATCHED

-0.68

cum

-0.67

hess

-0.65

ãĤ´ãĥ³

-0.65

 Stronghold

-0.65

POSITIVE LOGITS

 innocent

0.90

 unwitting

0.89

us

0.85

 insulting

0.83

 scapego

0.82

 undue

0.81

 oneself

0.81

 unnecessary

0.81

 people

0.80

 everybody

0.79

Activations Density 0.259%

verbs related to actions or commands

verbs related to creation and influence

No Comments

No Known Activations