INDEX

Explanations

words related to the conclusion or ending of a series or show

oai_token-act-pair · gpt-3.5-turbo

references to television series finales

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

artisan

-0.79

Cola

-0.77

hari

-0.74

uggage

-0.74

uge

-0.73

lying

-0.72

ums

-0.72

absolute

-0.71

urg

-0.71

elong

-0.70

POSITIVE LOGITS

 finale

0.99

[|

0.86

 airs

0.84

 episode

0.79

APR

0.79

 airing

0.79

 prematurely

0.78

 reel

0.77

arc

0.71

 aired

0.70

Activations Density 0.018%

words related to the conclusion or ending of a series or show

references to television series finales

No Comments

No Known Activations