INDEX

Explanations

instances where an action has occurred more than once

oai_token-act-pair · gpt-3.5-turbo

sentences or phrases that express repetition or occurrences happening multiple times

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

eries

-0.72

ibles

-0.67

uge

-0.63

quartered

-0.63

ãĥ¥

-0.61

ļéĨĴ

-0.60

tnc

-0.60

owed

-0.60

ãĤ´

-0.59

ords

-0.59

POSITIVE LOGITS

when

0.78

 amen

0.76

before

0.76

 sometime

0.73

imester

0.72

during

0.72

 whenever

0.70

 alas

0.70

 during

0.69

 thanks

0.69

Activations Density 0.428%

instances where an action has occurred more than once

sentences or phrases that express repetition or occurrences happening multiple times

No Comments

No Known Activations