INDEX

Explanations

titles or headings

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate the titles or names of articles and publications

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

Constructed

-0.61

chrom

-0.60

jen

-0.60

earances

-0.59

eda

-0.59

 bases

-0.59

 practiced

-0.59

ometers

-0.59

 base

-0.58

Hispanic

-0.58

POSITIVE LOGITS

"#

1.04

"<

0.88

selves

0.85

 Dreams

0.78

"$

0.76

 Beware

0.76

"_

0.75

 Operation

0.75

"(

0.75

 Miracle

0.74

Activations Density 0.065%

titles or headings

phrases that indicate the titles or names of articles and publications

No Comments

No Known Activations