INDEX

Explanations

numeric patterns with a dash between them

oai_token-act-pair · gpt-3.5-turbo

numeric representations related to quantity or measurements

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

alam

-0.75

gypt

-0.72

 wedge

-0.65

autical

-0.65

ournals

-0.65

 Miko

-0.63

 Rabb

-0.62

auder

-0.62

ucl

-0.61

Bird

-0.60

POSITIVE LOGITS

Shift

0.81

dden

0.77

VALUE

0.74

delete

0.68

zero

0.67

reset

0.66

limit

0.65

0.64

0.62

ffff

0.61

Activations Density 0.127%

numeric patterns with a dash between them

numeric representations related to quantity or measurements

No Comments

No Known Activations