INDEX

Explanations

instances where something could have happened or been different

oai_token-act-pair · gpt-3.5-turbo

phrases indicating hypothetical situations involving the phrase "could have been" or "would have been."

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

Deb

-0.73

ftime

-0.68

orate

-0.66

 suburbs

-0.65

Dj

-0.63

now

-0.61

 encl

-0.60

 stump

-0.59

oglu

-0.59

orns

-0.58

POSITIVE LOGITS

LOCK

0.80

sung

0.72

iscons

0.70

 Helpful

0.69

SPA

0.68

ypes

0.68

IU

0.68

anni

0.68

?????-

0.67

Activations Density 0.136%

instances where something could have happened or been different

phrases indicating hypothetical situations involving the phrase "could have been" or "would have been."

No Comments

No Known Activations