INDEX

Explanations

verbs preceded by "re" for words that reflect actions of reevaluation or reimagination

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "re-" indicating actions of repetition, restoration, or reconsideration

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 dying

-0.86

 wont

-0.85

 surviving

-0.80

 passing

-0.74

 proving

-0.72

 spilling

-0.70

-0.69

 losing

-0.68

 reproduction

-0.67

 bund

-0.67

POSITIVE LOGITS

activate

1.51

evaluate

1.49

create

1.41

establish

1.39

write

1.35

route

1.29

ify

1.26

join

1.24

enter

1.23

apply

1.21

Activations Density 0.054%

verbs preceded by "re" for words that reflect actions of reevaluation or reimagination

instances of the prefix "re-" indicating actions of repetition, restoration, or reconsideration

No Comments

No Known Activations

verbs preceded by "re" for words that reflect actions of reevaluation or reimagination

instances of the prefix "re-" indicating actions of repetition, restoration, or reconsideration

No Comments

No Known Activations