INDEX

Explanations

terms related to attackers and their actions

oai_token-act-pair · gpt-3.5-turbo

references to an "attacker" 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 @ 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

zl

-1.01

sit

-0.83

wheel

-0.82

Balt

-0.79

REE

-0.78

umph

-0.74

 Revival

-0.73

carb

-0.70

urgical

-0.70

QL

-0.68

POSITIVE LOGITS

 attackers

0.96

 attacker

0.91

 assailants

0.80

 attacked

0.78

 suspects

0.76

 beware

0.75

 rapist

0.73

 assailant

0.72

 attack

0.70

 intent

0.70

Activations Density 0.029%

terms related to attackers and their actions

references to an "attacker" in various contexts

No Comments

No Known Activations