INDEX

Explanations

brands, names, and initials

oai_token-act-pair · gpt-3.5-turbo

names, acronyms, and abbreviations related to music, culture, and entertainment

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

ï¸ı

-0.69

 tremend

-0.66

esson

-0.62

 trough

-0.61

sworth

-0.59

ãĤŃ

-0.58

ayson

-0.58

EY

-0.57

 exting

-0.57

ript

-0.56

POSITIVE LOGITS

 Moff

0.78

ylum

0.62

 Gadget

0.61

Osw

0.61

loo

0.59

 heals

0.59

Ha

0.57

 Cardiff

0.56

roo

0.56

 apologised

0.56

Activations Density 1.334%

brands, names, and initials

names, acronyms, and abbreviations related to music, culture, and entertainment

No Comments

No Known Activations