INDEX

Explanations

phrases indicating a call to action related to newsletters or emails for updates

oai_token-act-pair · gpt-3.5-turbo

instances of the word "and" and its variations in a context related to subscriptions or sign-ups

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

anu

-0.76

oldemort

-0.72

@@

-0.71

impl

-0.71

wagon

-0.68

umo

-0.67

establishment

-0.67

pex

-0.65

Ga

-0.65

dies

-0.65

POSITIVE LOGITS

 distribute

0.85

 enjoy

0.80

 click

0.79

 explore

0.78

 optionally

0.78

 hopefully

0.77

 then

0.76

 discuss

0.74

 bookmark

0.73

 refresh

0.73

Activations Density 0.239%

phrases indicating a call to action related to newsletters or emails for updates

instances of the word "and" and its variations in a context related to subscriptions or sign-ups

No Comments

No Known Activations