INDEX

Explanations

acronyms and capitalized terms, possibly related to government or formal organizations

oai_token-act-pair · gpt-3.5-turbo

occurrences of the term "con" 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 @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Beet

-0.76

-0.72

 Fired

-0.71

 Nicotine

-0.70

 Sapphire

-0.70

externalActionCode

-0.69

 Bulg

-0.67

ãĤ¶

-0.67

zeb

-0.66

iveness

-0.65

POSITIVE LOGITS

NECT

1.36

CEPT

0.92

CON

0.92

secut

0.92

TAIN

0.92

TERN

0.90

FORM

0.89

TEXT

0.88

STRUCT

0.87

bable

0.85

Activations Density 0.007%

acronyms and capitalized terms, possibly related to government or formal organizations

occurrences of the term "con" in various contexts

No Comments

No Known Activations