INDEX

Explanations

phrases indicating a call to action or a request to read something

oai_token-act-pair · gpt-3.5-turbo

phrases that encourage reading or referencing other content

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

akable

-0.84

hesda

-0.79

ongevity

-0.72

Operation

-0.71

ãĥŃ

-0.70

icable

-0.70

XL

-0.69

 srfAttach

-0.69

olute

-0.68

ngth

-0.67

POSITIVE LOGITS

 aloud

1.28

 newspapers

0.82

 papers

0.78

DragonMagazine

0.78

 cues

0.77

 drafts

0.75

 excerpts

0.75

 novels

0.74

papers

0.73

 mail

0.73

Activations Density 0.122%

phrases indicating a call to action or a request to read something

phrases that encourage reading or referencing other content

No Comments

No Known Activations