INDEX

Explanations

terms related to being searchable or upgradeable

oai_token-act-pair · gpt-3.5-turbo

instances of the word "able" in 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

itect

-0.85

krit

-0.76

erva

-0.75

urer

-0.71

herry

-0.69

akra

-0.67

aldi

-0.67

ospital

-0.66

arte

-0.66

ournal

-0.66

POSITIVE LOGITS

pedia

0.83

MpServer

0.69

atable

0.68

lihood

0.67

Redditor

0.66

ties

0.66

Ptr

0.65

 Haram

0.64

itarian

0.63

Characters

0.63

Activations Density 0.042%

terms related to being searchable or upgradeable

instances of the word "able" in various contexts

No Comments

No Known Activations