INDEX

Explanations

dates written in a specific format (full weekday, month, day, year)

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 unpre

-0.76

 Prelude

-0.69

 apprehension

-0.66

 bottleneck

-0.66

psy

-0.65

 manifold

-0.65

 revived

-0.64

 Metallic

-0.64

 doubling

-0.63

 flares

-0.63

POSITIVE LOGITS

isdom

1.34

alking

1.33

orst

1.33

idespread

1.30

restling

1.30

atson

1.29

esley

1.29

orthy

1.25

izards

1.25

olves

1.25

Activations Density 0.029%

dates written in a specific format (full weekday, month, day, year)

No Comments

No Known Activations