INDEX

Explanations

instances where the text is indicated as part of a larger concept or entity

oai_token-act-pair · gpt-3.5-turbo

phrases indicating parts of a whole or components of a larger system

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ares

-0.78

rams

-0.73

eers

-0.67

eer

-0.65

aires

-0.64

oons

-0.63

ients

-0.63

oths

-0.63

cats

-0.62

orks

-0.62

POSITIVE LOGITS

 reason

0.97

 impetus

0.81

 blame

0.74

 motivation

0.73

 explanation

0.72

hedral

0.72

©¶æ

0.69

 downside

0.69

 concern

0.68

 reasoning

0.67

Activations Density 0.123%

instances where the text is indicated as part of a larger concept or entity

phrases indicating parts of a whole or components of a larger system

No Comments

No Known Activations