INDEX

Explanations

legal or political terms

oai_token-act-pair · gpt-3.5-turbo

ellipses, indicating pauses or omissions in thought

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

»Ĵ

-0.82

 cradle

-0.73

 nerds

-0.67

ursed

-0.67

 asteroids

-0.66

ãĥ«

-0.64

ãĤ´ãĥ³

-0.64

selage

-0.64

åŃĲ

-0.64

 ageing

-0.64

POSITIVE LOGITS

etc

0.93

 âĢİ

0.87

police

0.82

emphasis

0.82

."

0.81

],

0.81

,"

0.81

among

0.81

they

0.80

gov

0.80

Activations Density 0.047%

legal or political terms

ellipses, indicating pauses or omissions in thought

No Comments

No Known Activations