INDEX

Explanations

the names of directors

oai_token-act-pair · gpt-3.5-turbo

instances of the word "by" indicating authorship or direction

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

cair

-0.90

opian

-0.89

ctic

-0.78

bard

-0.77

ndum

-0.77

asy

-0.75

ndra

-0.74

chwitz

-0.74

ulnerable

-0.73

tle

-0.73

POSITIVE LOGITS

 Mari

0.79

 Dame

0.77

 Wizards

0.77

 virtue

0.76

VID

0.74

 Julius

0.74

Pa

0.74

 Conan

0.73

 STATS

0.72

 Warner

0.72

Activations Density 0.098%

the names of directors

instances of the word "by" indicating authorship or direction

No Comments

No Known Activations