INDEX

Explanations

statements or calls related to being heard or making a voice heard

oai_token-act-pair · gpt-3.5-turbo

phrases that signify being acknowledged or listened to

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

 recons

-0.66

iets

-0.65

 contracted

-0.64

obs

-0.60

 antidepressants

-0.60

 revision

-0.59

ses

-0.58

 circumstance

-0.58

otomy

-0.57

tan

-0.57

POSITIVE LOGITS

 Hearing

0.84

 Voices

0.82

 Heard

0.79

ĸļ

0.78

 voices

0.77

 footsteps

0.76

 voic

0.75

 Hear

0.74

ropolitan

0.73

 hear

0.71

Activations Density 0.021%

statements or calls related to being heard or making a voice heard

phrases that signify being acknowledged or listened to

No Comments

No Known Activations