INDEX

Explanations

the word "there" occurring with high activation values

oai_token-act-pair · gpt-3.5-turbo

the repeated phrase "there" indicating emphasis on existence or presence in various contexts

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

shoot

-0.65

 Honour

-0.64

ONSORED

-0.62

 Heights

-0.59

ship

-0.58

uously

-0.57

tnc

-0.57

full

-0.56

isable

-0.56

Ã©e

-0.54

POSITIVE LOGITS

abouts

1.48

etheless

0.98

upon

0.95

fore

0.81

after

0.75

are

0.74

 aren

0.72

FORE

0.72

choes

0.71

enty

0.71

Activations Density 0.108%

the word "there" occurring with high activation values

the repeated phrase "there" indicating emphasis on existence or presence in various contexts

No Comments

No Known Activations