INDEX

Explanations

words related to destructive and criminal activities

oai_token-act-pair · gpt-3.5-turbo

conjunctions that connect multiple ideas or actions, particularly in contexts involving negative events or violence

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

yi

-0.78

lear

-0.75

hent

-0.73

fter

-0.72

meric

-0.68

Reviewer

-0.68

eric

-0.67

inks

-0.67

enne

-0.67

 Aware

-0.66

POSITIVE LOGITS

 injure

1.19

 stole

1.10

 destroying

1.05

 destroys

1.05

 threaten

1.04

 ruining

1.03

 jeopard

1.02

 vandal

1.02

 damaging

1.02

 livelihood

1.02

Activations Density 0.215%

words related to destructive and criminal activities

conjunctions that connect multiple ideas or actions, particularly in contexts involving negative events or violence

No Comments

No Known Activations