INDEX

Explanations

terms related to actions or instructions being provided

oai_token-act-pair · gpt-3.5-turbo

elements expressing opinions or assertions

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

emale

-0.81

 destro

-0.72

 submar

-0.70

anamo

-0.69

 Seym

-0.69

 Vaugh

-0.68

avorite

-0.67

 withd

-0.66

 occas

-0.65

aturdays

-0.64

POSITIVE LOGITS

 âĢº

0.73

lihood

0.63

 SHARES

0.62

 largeDownload

0.62

ables

0.59

ings

0.57

ye

0.57

 Vulkan

0.57

 Expand

0.57

 Hemp

0.56

Activations Density 0.131%

terms related to actions or instructions being provided

elements expressing opinions or assertions

No Comments

No Known Activations