INDEX

Explanations

proper nouns related to individuals involved in news stories

oai_token-act-pair · gpt-3.5-turbo

occurrences of the phrase "who was" followed by various descriptions or actions

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

zens

-0.64

 IMAGES

-0.63

ometers

-0.62

Footnote

-0.62

oke

-0.61

tarians

-0.61

ennett

-0.61

.>>

-0.61

lly

-0.60

ocate

-0.60

POSITIVE LOGITS

 formerly

1.08

 supposed

1.00

 previously

0.95

 originally

0.93

 conceived

0.90

 subsequently

0.89

 rumored

0.89

 instrumental

0.86

hes

0.86

 tasked

0.85

Activations Density 0.157%

proper nouns related to individuals involved in news stories

occurrences of the phrase "who was" followed by various descriptions or actions

No Comments

No Known Activations