INDEX

Explanations

mentions of the word "Hel" followed by a numerical value, particularly when the numerical value is high

oai_token-act-pair · gpt-3.5-turbo

references to the name "Hel" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

HAEL

-0.76

CLASSIFIED

-0.71

 Cunning

-0.69

EAR

-0.67

æĦ

-0.67

rers

-0.66

VALUE

-0.66

hyde

-0.65

parts

-0.65

çĶŁ

-0.65

POSITIVE LOGITS

mut

0.94

idon

0.91

mand

0.89

pless

0.88

abor

0.87

ios

0.86

ped

0.82

mes

0.80

ven

0.79

Hel

0.79

Activations Density 0.009%

mentions of the word "Hel" followed by a numerical value, particularly when the numerical value is high

references to the name "Hel" in various contexts

No Comments

No Known Activations