INDEX

Explanations

words related to authority, action, and controversial or impactful events

oai_token-act-pair · gpt-3.5-turbo

actions or states related to authorization and responsibility 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 @ 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

adra

-0.75

arat

-0.65

credit

-0.65

utra

-0.64

 Proposition

-0.63

apest

-0.63

yah

-0.63

é¾į

-0.63

ectar

-0.62

region

-0.62

POSITIVE LOGITS

 aback

0.85

rive

0.83

BY

0.77

ÃĥÃĤ

0.72

paren

0.72

 subur

0.72

river

0.71

own

0.71

 murd

0.70

 thereto

0.70

Activations Density 0.452%

words related to authority, action, and controversial or impactful events

actions or states related to authorization and responsibility in various contexts

No Comments

No Known Activations