INDEX

Explanations

terms related to tasks or activities

oai_token-act-pair · gpt-3.5-turbo

references to various tasks and chores

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

 Freddie

-0.71

ounded

-0.68

soType

-0.67

brace

-0.66

agogue

-0.64

nat

-0.64

 Graham

-0.63

spread

-0.63

minent

-0.61

ohn

-0.61

POSITIVE LOGITS

 tasks

3.57

 task

2.36

 chores

2.29

task

1.97

Task

1.85

 duties

1.70

 responsibilities

1.62

 assignments

1.60

 Task

1.57

asks

1.53

Activations Density 0.014%

terms related to tasks or activities

references to various tasks and chores

No Comments

No Known Activations