INDEX

Explanations

the word "Raj" with varying levels of intensity

oai_token-act-pair · gpt-3.5-turbo

mentions of a specific name or title

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

 Scotia

-0.65

VALUE

-0.64

 Mobil

-0.64

 Shards

-0.61

 magnification

-0.59

 disinfect

-0.59

 Takeru

-0.58

ancial

-0.58

 Ethiop

-0.58

backer

-0.58

POSITIVE LOGITS

ournal

1.54

ournals

1.04

ean

1.02

itsu

1.00

arro

0.92

ewski

0.86

olla

0.84

eh

0.83

unction

0.82

ohn

0.82

Activations Density 0.021%

the word "Raj" with varying levels of intensity

mentions of a specific name or title

No Comments

No Known Activations