INDEX

Explanations

phrases contrasting two different concepts

oai_token-act-pair · gpt-3.5-turbo

phrases that convey contrast or exceptions

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

uter

-0.81

velt

-0.76

unction

-0.74

ct

-0.74

ury

-0.72

tty

-0.70

ory

-0.69

ctor

-0.69

UD

-0.69

asons

-0.69

POSITIVE LOGITS

 suffice

0.92

 nevertheless

0.91

nor

0.88

 rather

0.88

 alas

0.85

 nonetheless

0.83

hey

0.79

 fortunately

0.76

 luckily

0.76

 merely

0.74

Activations Density 0.125%

phrases contrasting two different concepts

phrases that convey contrast or exceptions

No Comments

No Known Activations