INDEX

Explanations

names of individuals giving speeches or addresses

oai_token-act-pair · gpt-3.5-turbo

instances of the word "address" and related terms indicating speeches or formal communications

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

 monitoring

-0.68

 downstream

-0.68

 increment

-0.62

 collabor

-0.61

lia

-0.61

 perception

-0.61

 experimenting

-0.61

 tolerated

-0.60

 endanger

-0.60

 tolerate

-0.60

POSITIVE LOGITS

 speeches

0.99

itone

0.86

 aloud

0.86

esta

0.78

alla

0.78

 loudly

0.77

 Speech

0.75

 keynote

0.74

aug

0.74

 commencement

0.72

Activations Density 0.118%

names of individuals giving speeches or addresses

instances of the word "address" and related terms indicating speeches or formal communications

No Comments

No Known Activations