INDEX

Explanations

references to superlatives or extremes, such as "best", "largest", "greatest", and "most"

oai_token-act-pair · gpt-3.5-turbo

the occurrence of the word "in" within various contexts

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

EngineDebug

-0.80

DragonMagazine

-0.79

ido

-0.69

mil

-0.68

oother

-0.68

-0.67

OTH

-0.67

bos

-0.66

ango

-0.66

REL

-0.63

POSITIVE LOGITS

ordinate

0.99

clusively

0.93

animate

0.92

efficiency

0.90

 vitro

0.89

patient

0.89

 lieu

0.86

roads

0.85

 regards

0.84

clusions

0.84

Activations Density 0.138%

references to superlatives or extremes, such as "best", "largest", "greatest", and "most"

the occurrence of the word "in" within various contexts

No Comments

No Known Activations