INDEX

Explanations

words related to having two of something or involving a pair

oai_token-act-pair · gpt-3.5-turbo

references to new entities or categories

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

=-=-=-=-=-=-=-=-

-0.72

antry

-0.70

 Petra

-0.70

someone

-0.68

 FANTASY

-0.68

 Abdel

-0.68

uddin

-0.68

rology

-0.67

 Emerson

-0.66

 Beir

-0.65

POSITIVE LOGITS

 thirds

0.88

 apiece

0.86

 halves

0.85

 streams

0.75

 brothers

0.75

iliated

0.74

handled

0.74

 sisters

0.74

TDs

0.72

 totaling

0.71

Activations Density 0.227%

words related to having two of something or involving a pair

references to new entities or categories

No Comments

No Known Activations