INDEX

Explanations

letter, number, and symbol sequences

oai_token-act-pair · gpt-3.5-turbo

terms associated with data values or metrics

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 debugging

-0.76

 scrut

-0.74

 princ

-0.73

 upkeep

-0.69

 trave

-0.69

 stuffing

-0.69

 closures

-0.69

 refres

-0.68

 descript

-0.68

 derog

-0.67

POSITIVE LOGITS

zx

1.33

ZX

1.28

FK

1.24

NK

1.21

kl

1.19

kj

1.18

1.17

Uk

1.16

qq

1.16

Gh

1.15

Activations Density 0.076%

letter, number, and symbol sequences

terms associated with data values or metrics

No Comments

No Known Activations