INDEX

Explanations

phrases related to enabling actions or functionalities

oai_token-act-pair · gpt-3.5-turbo

phrases indicating the functionality or capabilities of tools and applications

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

boy

-0.79

ta

-0.71

borough

-0.70

bil

-0.67

wa

-0.67

xon

-0.64

source

-0.64

bons

-0.64

tone

-0.60

town

-0.59

POSITIVE LOGITS

Reviewer

0.98

geries

0.90

Allows

0.83

uces

0.77

us

0.77

hift

0.73

 withdrawals

0.72

ibaba

0.72

bidden

0.71

ences

0.71

Activations Density 0.043%

phrases related to enabling actions or functionalities

phrases indicating the functionality or capabilities of tools and applications

No Comments

No Known Activations