INDEX

Explanations

phrases related to explaining or confirming information

oai_token-act-pair · gpt-3.5-turbo

instances of the word "to" indicating explanations or directives

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

 downtime

-0.69

 distractions

-0.64

 botched

-0.61

 wasted

-0.59

 mobility

-0.59

 coupled

-0.57

 exposures

-0.56

 pitted

-0.56

 wasting

-0.56

 simultane

-0.56

POSITIVE LOGITS

wered

1.12

ilet

1.09

pless

0.99

ggles

0.97

othy

0.96

ads

0.89

 satisfy

0.86

asted

0.86

adies

0.85

psy

0.84

Activations Density 0.273%

phrases related to explaining or confirming information

instances of the word "to" indicating explanations or directives

No Comments

No Known Activations