INDEX

Explanations

words related to newsletters or email subscriptions

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

references to newsletters and subscription-related terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_post

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_post

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

Dialogue

-0.66

 Strauss

-0.64

bered

-0.61

 derog

-0.61

 heel

-0.60

ãĥ´

-0.60

 sled

-0.59

 blame

-0.59

WC

-0.58

joy

-0.58

POSITIVE LOGITS

 Upgrade

0.84

LET

0.78

letter

0.77

 Sign

0.75

 Subscribe

0.75

 Enter

0.72

 subscribers

0.69

letters

0.69

 Gifts

0.68

 Understand

0.67

Activations Density 0.027%

words related to newsletters or email subscriptions

references to newsletters and subscription-related terms

No Comments

No Known Activations