INDEX

Explanations

phrases related to permission, authorization or enabling actions

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate potential or capability, particularly those beginning with "to."

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

quickShipAvailable

-0.80

furt

-0.73

lights

-0.68

ELD

-0.66

 meant

-0.65

mails

-0.64

 Coff

-0.64

ĪĴ

-0.63

listed

-0.62

 Explan

-0.62

POSITIVE LOGITS

 communicate

1.19

 concentrate

1.15

 survive

1.14

 capitalize

1.13

 negotiate

1.13

 differentiate

1.13

 pursue

1.11

 utilize

1.11

 manipulate

1.10

 achieve

1.10

Activations Density 0.123%

phrases related to permission, authorization or enabling actions

phrases that indicate potential or capability, particularly those beginning with "to."

No Comments

No Known Activations