INDEX

Explanations

instances where something is possible or scenarios where something can be done

oai_token-act-pair · gpt-3.5-turbo

instances where the word "could" indicates potential or ability

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

cedented

-0.72

furt

-0.69

balls

-0.65

ente

-0.61

holder

-0.60

houses

-0.60

 Yards

-0.59

ieu

-0.59

cloth

-0.58

Ing

-0.58

POSITIVE LOGITS

 feas

1.11

 afford

1.01

 conce

0.97

 foresee

0.91

adian

0.90

berra

0.89

nt

0.87

 conclud

0.85

've

0.84

 possibly

0.84

Activations Density 0.085%

instances where something is possible or scenarios where something can be done

instances where the word "could" indicates potential or ability

No Comments

No Known Activations