INDEX

Explanations

texts containing notes or messages left for someone

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "note."

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

eatures

-0.75

emouth

-0.74

ãĥĩ

-0.73

rates

-0.70

heed

-0.68

anooga

-0.68

namese

-0.67

imates

-0.67

imated

-0.67

slaught

-0.66

POSITIVE LOGITS

BOOK

1.12

book

1.04

books

1.02

 Keeper

0.93

ably

0.91

notes

0.86

manship

0.78

 note

0.74

Pad

0.72

 notes

0.70

Activations Density 0.016%

texts containing notes or messages left for someone

occurrences of the word "note."

No Comments

No Known Activations