INDEX

Explanations

legal or justice-related terms, specifically focusing on the concept of conviction

oai_token-act-pair · gpt-3.5-turbo

references to legal convictions

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

ummer

-0.78

hod

-0.76

ebus

-0.74

gm

-0.74

ÃŁ

-0.71

hens

-0.70

kef

-0.69

berman

-0.68

pee

-0.67

 nutrit

-0.66

POSITIVE LOGITS

 convictions

1.12

 conviction

1.03

eering

0.86

icts

0.78

iction

0.75

alties

0.75

enance

0.75

 warrant

0.72

 convicted

0.71

 overturned

0.71

Activations Density 0.009%

legal or justice-related terms, specifically focusing on the concept of conviction

references to legal convictions

No Comments

No Known Activations