INDEX

Explanations

pieces of text mentioning information or details

oai_token-act-pair · gpt-3.5-turbo

references to information and informational content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 vain

-0.69

 Sabha

-0.65

 tense

-0.65

jug

-0.64

negie

-0.63

deen

-0.62

 tranquil

-0.60

ple

-0.60

ood

-0.58

nered

-0.58

POSITIVE LOGITS

 info

1.20

information

0.89

 Info

0.86

ioch

0.79

 INFORMATION

0.78

escription

0.78

imei

0.77

info

0.77

ãĤ±

0.75

Info

0.73

Activations Density 0.009%

pieces of text mentioning information or details

references to information and informational content

No Comments

No Known Activations