INDEX

Explanations

words related to specific entities or events, possibly related to news articles or stories

oai_token-act-pair · gpt-3.5-turbo

occurrences of a specific symbol or character related to entities or concepts in text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 limb

-0.76

 writ

-0.75

 myster

-0.73

 vulner

-0.73

 Beir

-0.68

 Vaugh

-0.66

bun

-0.65

 conj

-0.64

 trainers

-0.64

Vog

-0.64

POSITIVE LOGITS

ï¸ı

1.29

vernment

1.28

ËĪ

1.19

Ô

1.16

lean

1.09

ãĥĥãĥī

1.06

SpaceEngineers

1.02

ï¸

1.02

ðĿ

1.02

âĹ¼

0.99

Activations Density 0.042%

words related to specific entities or events, possibly related to news articles or stories

occurrences of a specific symbol or character related to entities or concepts in text

No Comments

No Known Activations