INDEX

Explanations

statements of fact or correctness

oai_token-act-pair · gpt-3.5-turbo

statements about correctness or truth

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

andise

-0.80

ogether

-0.77

VERTISEMENT

-0.76

icipated

-0.75

heastern

-0.73

 actionGroup

-0.71

berus

-0.71

FTWARE

-0.70

ð

-0.70

letal

-0.68

POSITIVE LOGITS

 referring

1.44

 adamant

1.36

 quoting

1.23

 correct

1.22

 proposing

1.20

 arguing

1.19

 pessimistic

1.17

 mistaken

1.17

 skeptical

1.16

 aware

1.15

Activations Density 0.270%

statements of fact or correctness

statements about correctness or truth

No Comments

No Known Activations