INDEX

Explanations

names containing the string "rov"

oai_token-act-pair · gpt-3.5-turbo

references to a specific individual or entity associated with authority or governance

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 amber

-0.59

 Tsukuyomi

-0.58

 America

-0.58

 judgment

-0.57

Mis

-0.57

HMS

-0.57

 judgement

-0.56

mas

-0.55

 tails

-0.55

 Dock

-0.55

POSITIVE LOGITS

rov

4.77

riv

1.27

sov

1.20

chev

1.19

akov

1.08

reb

1.08

henko

1.08

iov

1.05

rog

1.05

rox

1.05

Activations Density 0.011%

names containing the string "rov"

references to a specific individual or entity associated with authority or governance

No Comments

No Known Activations