INDEX

Explanations

words related to TV or radio broadcasts

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "aired."

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

llan

-0.88

illard

-0.87

ansson

-0.77

ever

-0.75

ledge

-0.73

opher

-0.71

 Fraz

-0.71

acqu

-0.69

anova

-0.69

idget

-0.69

POSITIVE LOGITS

 aired

0.88

 airing

0.87

ONY

0.86

waves

0.79

ãĥĺ

0.76

ÃįÃį

0.75

NPR

0.75

BBC

0.74

Ô

0.73

OHN

0.73

Activations Density 0.017%

words related to TV or radio broadcasts

occurrences of the word "aired."

No Comments

No Known Activations