INDEX

Explanations

words related to operations or cooperating with others

oai_token-act-pair · gpt-3.5-turbo

instances of the word "oper" or related variations indicating various operations or processes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 vengeance

-0.70

 Primordial

-0.69

 fateful

-0.63

 haste

-0.60

ĪĴ

-0.59

fix

-0.59

 Cere

-0.59

definition

-0.57

+++

-0.57

 kcal

-0.56

POSITIVE LOGITS

oper

1.04

ament

0.96

aquin

0.90

atives

0.87

ating

0.81

ates

0.80

atis

0.80

merce

0.80

ational

0.80

inson

0.79

Activations Density 0.008%

words related to operations or cooperating with others

instances of the word "oper" or related variations indicating various operations or processes

No Comments

No Known Activations