INDEX

Explanations

dates and locations expressed as coordinates like 'Latitudinal and Longitudinal'

oai_token-act-pair · gpt-3.5-turbo

phrases starting with "By"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

kered

-0.77

hes

-0.58

vable

-0.55

essage

-0.55

illon

-0.54

 stuck

-0.53

itans

-0.53

 unemployment

-0.53

ady

-0.53

 worth

-0.52

POSITIVE LOGITS

By

3.06

By

2.25

BY

1.64

BY

1.44

by

1.42

 Through

1.42

To

1.25

Via

1.23

For

1.22

 With

1.21

Activations Density 0.011%

dates and locations expressed as coordinates like 'Latitudinal and Longitudinal'

phrases starting with "By"

No Comments

No Known Activations