INDEX

Explanations

references to early time, like early in the day or leaving early

oai_token-act-pair · gpt-3.5-turbo

instances of the word "early."

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

uyomi

-0.78

retched

-0.77

ortium

-0.75

utenberg

-0.74

ittee

-0.74

alach

-0.71

 Uriel

-0.68

edIn

-0.68

ĸļ

-0.67

shall

-0.66

POSITIVE LOGITS

 adop

0.91

 stages

0.83

 morning

0.79

 onset

0.77

 mornings

0.75

 drafts

0.73

 bird

0.73

 afternoon

0.72

bird

0.72

 enough

0.72

Activations Density 0.025%

references to early time, like early in the day or leaving early

instances of the word "early."

No Comments

No Known Activations