INDEX

Explanations

evidence-related terms and phrases

oai_token-act-pair · gpt-3.5-turbo

terms related to various aspects of evidence in criminal cases

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

•Evidence

No Comments

Negative Logits

ategory

-0.77

orks

-0.75

awar

-0.74

yright

-0.69

lisher

-0.68

TOR

-0.66

Hop

-0.66

 Wonderful

-0.66

pop

-0.65

idays

-0.65

POSITIVE LOGITS

 tampering

1.03

 evidence

0.94

evidence

0.91

abilia

0.90

intel

0.81

 Evidence

0.80

 testifying

0.77

glass

0.75

uments

0.73

 incrim

0.73

Activations Density 0.036%

evidence-related terms and phrases

terms related to various aspects of evidence in criminal cases

No Comments

No Known Activations