INDEX

Explanations

references to court justices and legal proceedings through specific words and phrases

oai_token-act-pair · gpt-3.5-turbo

references to justices in a legal context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ettel

-0.76

ahon

-0.71

Ã§

-0.68

eworld

-0.67

uilt

-0.67

é¾

-0.66

 Kurd

-0.66

 Azerb

-0.64

ivation

-0.63

 Caption

-0.63

POSITIVE LOGITS

 justices

1.20

ervative

0.97

ervatives

0.86

 Scalia

0.85

otle

0.81

Justice

0.81

hip

0.80

justice

0.80

 laureate

0.77

hips

0.76

Activations Density 0.015%

references to court justices and legal proceedings through specific words and phrases

references to justices in a legal context

No Comments

No Known Activations