INDEX

Explanations

ranks and titles followed by names or specific positions

oai_token-act-pair · gpt-3.5-turbo

references to military and governmental titles and positions

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

zsche

-1.02

orrow

-0.75

rawdownloadcloneembedreportprint

-0.68

izu

-0.67

oho

-0.66

achus

-0.66

akespeare

-0.65

erella

-0.65

erest

-0.64

ichever

-0.63

POSITIVE LOGITS

 Surge

0.81

 Colonel

0.78

utenant

0.77

 Alvin

0.73

 Morales

0.72

 Commander

0.71

Interested

0.71

Lt

0.71

Col

0.66

 Orient

0.65

Activations Density 0.054%

ranks and titles followed by names or specific positions

references to military and governmental titles and positions

No Comments

No Known Activations