INDEX

Explanations

terms related to the concept of being "based on" something

oai_token-act-pair · gpt-3.5-turbo

phrases that include the word "based."

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

apes

-0.77

adobe

-0.75

 Appears

-0.69

acks

-0.67

å¥³

-0.67

apers

-0.66

ugh

-0.65

antha

-0.65

 Mouth

-0.64

ests

-0.63

POSITIVE LOGITS

 loosely

0.98

 upon

0.95

 solely

0.88

 entirely

0.78

 principally

0.75

illac

0.74

 purely

0.71

 atop

0.70

 primarily

0.69

 centrif

0.67

Activations Density 0.047%

terms related to the concept of being "based on" something

phrases that include the word "based."

No Comments

No Known Activations