INDEX

Explanations

phrases related to inquiries or investigations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "ires."

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

 sneak

-0.68

 beating

-0.68

 Consolid

-0.66

 sneaking

-0.64

 lump

-0.63

 correctly

-0.63

 context

-0.61

 placement

-0.61

 sucker

-0.60

 Sloan

-0.60

POSITIVE LOGITS

ires

4.81

ire

2.46

iring

2.36

ired

2.16

IRE

1.75

iries

1.71

irable

1.60

IRED

1.52

iry

1.51

irement

1.48

Activations Density 0.006%

phrases related to inquiries or investigations

instances of the word "ires."

No Comments

No Known Activations