INDEX

Explanations

phrases or names starting with the letters 'ph'

oai_token-act-pair · gpt-3.5-turbo

abbreviations or references to specific concepts or entities, particularly those starting with 'E'

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

REL

-0.67

xus

-0.64

caps

-0.60

ensitivity

-0.58

uties

-0.58

 ingred

-0.58

channelAvailability

-0.57

\/\/

-0.57

 kittens

-0.56

ahoo

-0.56

POSITIVE LOGITS

ilon

0.89

anasia

0.77

Reloaded

0.76

tainment

0.73

vironment

0.73

 coli

0.70

ascript

0.70

hardt

0.70

olon

0.68

pecially

0.68

Activations Density 0.076%

phrases or names starting with the letters 'ph'

abbreviations or references to specific concepts or entities, particularly those starting with 'E'

No Comments

No Known Activations