INDEX

Explanations

proper nouns or titles related to public speakers and conferences

oai_token-act-pair · gpt-3.5-turbo

references to speakers at events

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

 Soldiers

-0.77

 Pradesh

-0.75

hens

-0.73

ster

-0.67

uilt

-0.66

 Mines

-0.65

 Territories

-0.64

 Chart

-0.63

 Devils

-0.62

YP

-0.61

POSITIVE LOGITS

 speakers

1.18

 speaker

1.09

phone

0.90

voice

0.88

 voices

0.85

hip

0.83

 Emin

0.77

 presentations

0.75

speech

0.75

ainer

0.74

Activations Density 0.028%

proper nouns or titles related to public speakers and conferences

references to speakers at events

No Comments

No Known Activations