INDEX

Explanations

words related to errors or defects that can lead to negative outcomes

oai_token-act-pair · gpt-3.5-turbo

words associated with flaws and vulnerabilities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

ourses

-0.68

 administrations

-0.66

arie

-0.64

 Presidents

-0.62

 Boards

-0.62

 Tanks

-0.61

 Officers

-0.61

 Articles

-0.61

achers

-0.60

aukee

-0.60

POSITIVE LOGITS

 flaw

4.10

 flaws

1.97

 defect

1.50

 glitch

1.47

 loophole

1.45

 drawback

1.32

 fault

1.32

 weakness

1.24

 mistake

1.22

 vulnerability

1.16

Activations Density 0.011%

words related to errors or defects that can lead to negative outcomes

words associated with flaws and vulnerabilities

No Comments

No Known Activations