INDEX

Explanations

words ending in "-hes" or "-hes" itself

oai_token-act-pair · gpt-3.5-turbo

the presence of the word "hes" in various contexts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 congress

-0.64

Reviewer

-0.62

 planning

-0.62

 coff

-0.62

CAL

-0.61

recomm

-0.60

 collusion

-0.60

 link

-0.60

 reference

-0.59

 Doctors

-0.59

POSITIVE LOGITS

hes

1.18

apeake

1.18

creen

1.01

omen

0.97

terday

0.92

borough

0.91

hens

0.89

ervative

0.89

 ashore

0.88

piration

0.88

Activations Density 0.006%

words ending in "-hes" or "-hes" itself

the presence of the word "hes" in various contexts

No Comments

No Known Activations