INDEX

Explanations

phrases describing something as being completely allowed or appropriate

oai_token-act-pair · gpt-3.5-turbo

phrases related to the concept of perfection or acceptability

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

anwhile

-0.83

NetMessage

-0.80

ulative

-0.76

hyde

-0.71

jri

-0.68

mere

-0.68

worldly

-0.68

Downloadha

-0.67

ulton

-0.66

uese

-0.65

POSITIVE LOGITS

 suited

1.02

 aligned

0.92

 timed

0.82

 positioned

0.80

fit

0.74

 tuned

0.74

 encaps

0.73

@@

0.73

uner

0.73

 fitted

0.72

Activations Density 0.010%

phrases describing something as being completely allowed or appropriate

phrases related to the concept of perfection or acceptability

No Comments

No Known Activations