INDEX

Explanations

references to authority figures and law enforcement personnel

oai_token-act-pair · gpt-3.5-turbo

references to authority figures and security services

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ãĥ´

-0.75

£ı

-0.69

igue

-0.68

USS

-0.68

ausible

-0.68

ãĥĥãĥī

-0.67

fter

-0.67

RT

-0.64

Ñı

-0.64

phabet

-0.63

POSITIVE LOGITS

 condesc

1.01

 swoop

0.99

 beck

0.98

 intervene

0.98

 forbid

0.97

 kindly

0.97

 reprim

0.95

 pressuring

0.95

 evict

0.94

 disappro

0.93

Activations Density 0.725%

references to authority figures and law enforcement personnel

references to authority figures and security services

No Comments

No Known Activations