INDEX

Explanations

phrases related to instructions or directives

oai_token-act-pair · gpt-3.5-turbo

phrases emphasizing the concept of "keeping" or maintaining something

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

ĪĴ

-0.77

bernatorial

-0.74

ALLY

-0.74

haps

-0.70

DIT

-0.69

BAT

-0.69

esar

-0.66

KA

-0.65

 PRESIDENT

-0.65

JA

-0.65

POSITIVE LOGITS

 afloat

1.06

 footing

0.84

 alive

0.78

 wraps

0.77

 composure

0.74

readable

0.72

lid

0.70

 momentum

0.70

 tabs

0.69

 reins

0.69

Activations Density 0.108%

phrases related to instructions or directives

phrases emphasizing the concept of "keeping" or maintaining something

No Comments

No Known Activations