INDEX

Explanations

the word "Series" or phrases related to series finales

oai_token-act-pair · gpt-3.5-turbo

mentions of specific television series

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

zie

-0.77

orget

-0.72

zan

-0.69

rian

-0.68

selves

-0.68

attled

-0.67

lig

-0.67

ria

-0.66

arest

-0.66

bell

-0.65

POSITIVE LOGITS

 finale

0.99

manship

0.92

 regulars

0.79

hips

0.76

 Reborn

0.75

peak

0.73

pace

0.73

hip

0.73

ĸļ

0.72

ends

0.72

Activations Density 0.037%

the word "Series" or phrases related to series finales

mentions of specific television series

No Comments

No Known Activations