INDEX

Explanations

the word "thing."

oai_token-act-pair · gpt-3.5-turbo

references to "one thing" or similar phrases emphasizing a key point or idea

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

cled

-0.81

inav

-0.79

ONSORED

-0.77

DOM

-0.76

ãĥ¥

-0.72

imen

-0.67

EGIN

-0.66

NAS

-0.65

DOS

-0.65

cling

-0.65

POSITIVE LOGITS

 Valiant

0.82

 happens

0.82

iverse

0.78

 happening

0.74

 happened

0.74

 separates

0.74

 kicker

0.71

 counts

0.69

rued

0.68

 Subtle

0.67

Activations Density 0.028%

the word "thing."

references to "one thing" or similar phrases emphasizing a key point or idea

No Comments

No Known Activations