INDEX

Explanations

phrases related to a group or collection of items

oai_token-act-pair · gpt-3.5-turbo

references to groups of items or collections

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

plays

-0.70

ruption

-0.69

Ped

-0.67

pires

-0.67

blems

-0.65

 poles

-0.65

 Episcopal

-0.64

mington

-0.62

thy

-0.61

PLIED

-0.61

POSITIVE LOGITS

 batch

1.30

 batches

1.21

batch

0.94

mates

0.91

olean

0.81

DragonMagazine

0.74

Dispatch

0.74

fuck

0.73

 mates

0.70

bugs

0.69

Activations Density 0.004%

phrases related to a group or collection of items

references to groups of items or collections

No Comments

No Known Activations