INDEX

Explanations

words related to attacks or threatening situations, especially involving an attacker

oai_token-act-pair · gpt-3.5-turbo

references to an individual or group involved in aggressive actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

zl

-0.97

wheel

-0.83

sit

-0.71

 Hemp

-0.70

 Smoking

-0.64

tz

-0.64

 Mineral

-0.63

carb

-0.63

aver

-0.62

 Matter

-0.61

POSITIVE LOGITS

 attacker

1.38

 attackers

1.34

 assailant

0.98

 attack

0.98

attack

0.97

 assailants

0.97

 attacked

0.93

 intruder

0.90

 attacks

0.82

 Attack

0.80

Activations Density 0.005%

words related to attacks or threatening situations, especially involving an attacker

references to an individual or group involved in aggressive actions

No Comments

No Known Activations