INDEX

Explanations

abbreviations, trademarks, and proper nouns

oai_token-act-pair · gpt-3.5-turbo

distinct letters or combinations of letters in text

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

 underrated

-0.66

theless

-0.64

 Admir

-0.62

 Cornwall

-0.59

Xer

-0.57

 takeoff

-0.57

 Vers

-0.57

icut

-0.57

 compliment

-0.56

advertisement

-0.55

POSITIVE LOGITS

cific

0.85

letal

0.83

nikov

0.81

uler

0.79

Ó

0.77

ificant

0.77

pta

0.76

raper

0.75

atism

0.74

pport

0.73

Activations Density 0.316%

abbreviations, trademarks, and proper nouns

distinct letters or combinations of letters in text

No Comments

No Known Activations