INDEX

Explanations

words related to computer code or errors

oai_token-act-pair · gpt-3.5-turbo

the presence of the word "None" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

illon

-0.76

romy

-0.73

 widest

-0.71

rod

-0.68

cum

-0.67

roc

-0.63

Gall

-0.63

 multiplying

-0.62

cru

-0.62

pg

-0.62

POSITIVE LOGITS

theless

1.01

 Detected

0.87

essee

0.81

 Died

0.77

 endif

0.76

ecided

0.75

orem

0.72

None

0.72

etheless

0.70

icularly

0.68

Activations Density 0.008%

words related to computer code or errors

the presence of the word "None" in various contexts

No Comments

No Known Activations