INDEX

Explanations

references to recurring roles or themes in various contexts

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

phrases related to things happening repeatedly or appearing frequently

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ablishment

-0.87

chens

-0.79

eah

-0.74

orgetown

-0.70

cellence

-0.69

ugen

-0.69

imb

-0.68

okin

-0.68

toc

-0.68

ull

-0.67

POSITIVE LOGITS

 recurring

1.38

 recurrent

1.03

 nightmares

1.02

 occurrences

0.93

 themes

0.93

 episodes

0.91

 installments

0.90

 bouts

0.89

 motif

0.88

interstitial

0.85

Activations Density 0.004%

references to recurring roles or themes in various contexts

phrases related to things happening repeatedly or appearing frequently

No Comments

No Known Activations