INDEX

Explanations

German and European words

oai_token-act-pair · gpt-3.5-turbo

the word "nicht," indicating a focus on negation or denial

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 breastfeeding

-0.67

 reciproc

-0.65

 houses

-0.64

table

-0.64

Squ

-0.63

Dug

-0.61

MIL

-0.61

 multiplier

-0.60

 table

-0.59

 breast

-0.59

POSITIVE LOGITS

icht

4.52

ich

1.19

mberg

1.12

cht

1.11

beit

1.09

uchs

1.09

zik

1.07

ingen

1.06

ettel

1.06

wagen

1.05

Activations Density 0.012%

German and European words

the word "nicht," indicating a focus on negation or denial

No Comments

No Known Activations