INDEX

Explanations

names of news agencies or press organizations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Associated" and its variations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 WARN

-0.81

isan

-0.72

ysis

-0.69

fi

-0.68

 Beta

-0.66

bor

-0.64

ILLE

-0.63

jar

-0.62

gger

-0.62

tan

-0.61

POSITIVE LOGITS

Associated

0.98

 reperto

0.95

agascar

0.85

Images

0.80

onomous

0.74

jriwal

0.74

 waivers

0.72

nesses

0.71

ocalypse

0.71

interstitial

0.70

Activations Density 0.013%

names of news agencies or press organizations

instances of the word "Associated" and its variations

No Comments

No Known Activations