INDEX

Explanations

proper nouns such as names of people, places, and organizations

oai_token-act-pair · gpt-3.5-turbo

references to notable people, places, or organizations

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

¿½

-0.99

 Ezek

-0.73

 Mehran

-0.61

EMP

-0.60

 expansive

-0.60

 Mell

-0.60

COMPLE

-0.59

 nascent

-0.58

 Naples

-0.58

asury

-0.58

POSITIVE LOGITS

 sucks

1.30

ain

1.22

!!!

1.07

 doesnt

1.06

?!

1.06

 hates

1.05

!!!!

1.04

!!

0.97

!?

0.96

huh

0.96

Activations Density 0.646%

proper nouns such as names of people, places, and organizations

references to notable people, places, or organizations

No Comments

No Known Activations