INDEX

Explanations

words related to specific proper nouns, potentially names or places

oai_token-act-pair · gpt-3.5-turbo

abbreviations or acronyms, particularly those related to academic roles or organizations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

PF

-0.75

 gears

-0.70

igham

-0.69

ĸļ

-0.65

ĨĴ

-0.65

 magnification

-0.64

 peas

-0.61

 constants

-0.60

ibaba

-0.59

 levers

-0.59

POSITIVE LOGITS

ause

0.73

vantage

0.69

aline

0.68

ainment

0.66

anamo

0.66

ucl

0.65

alty

0.64

ublic

0.62

ITION

0.62

vernment

0.61

Activations Density 0.125%

words related to specific proper nouns, potentially names or places

abbreviations or acronyms, particularly those related to academic roles or organizations

No Comments

No Known Activations