INDEX

Explanations

URLs and file paths

oai_token-act-pair · gpt-3.5-turbo

patterns related to URLs or web links

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

 creep

-0.71

 apologised

-0.68

 creeping

-0.66

 wards

-0.66

 Armageddon

-0.65

 angry

-0.62

 tight

-0.61

 Weir

-0.61

 Baal

-0.61

Ago

-0.61

POSITIVE LOGITS

index

1.12

photos

1.09

upload

1.04

pol

1.01

detail

1.01

src

1.01

comments

0.99

archives

0.99

uploads

0.99

Accessory

0.98

Activations Density 0.041%

URLs and file paths

patterns related to URLs or web links

No Comments

No Known Activations