INDEX

Explanations

numerical values ending with '1' and sections with phrases that involve placing human lives at risk

oai_token-act-pair · gpt-3.5-turbo

Arabic characters and symbols related to text encoding issues

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

nect

-0.97

NetMessage

-0.85

istically

-0.84

*/(

-0.83

onomy

-0.80

anooga

-0.80

fman

-0.79

glers

-0.77

 hitch

-0.77

inately

-0.76

POSITIVE LOGITS

³

0.89

×ª

0.84

à¥

0.84

ÙĦ

0.82

Į

0.80

ÙĨ

0.78

——

0.77

¸

0.77

Ö¼

0.77

´

0.77

Activations Density 0.009%

numerical values ending with '1' and sections with phrases that involve placing human lives at risk

Arabic characters and symbols related to text encoding issues

No Comments

No Known Activations