INDEX

Explanations

phrases related to titles of books, movies, and journals

oai_token-act-pair · gpt-3.5-turbo

phrases that reference the structure or content of books, movies, or other media

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

rouse

-0.78

 sympath

-0.74

 advis

-0.74

 eater

-0.73

 balcon

-0.72

 versa

-0.70

 offending

-0.70

entimes

-0.70

 jaws

-0.69

 brake

-0.69

POSITIVE LOGITS

 Names

1.09

 Excellence

1.03

 Life

1.03

 Them

1.03

 Everyday

1.02

 Yourself

1.02

 Lies

1.00

 Violence

1.00

 Tomorrow

1.00

 Madness

0.99

Activations Density 0.180%

phrases related to titles of books, movies, and journals

phrases that reference the structure or content of books, movies, or other media

No Comments

No Known Activations