INDEX

Explanations

phrases related to evidence or proof

oai_token-act-pair · gpt-3.5-turbo

negative phrases related to validity or evidence

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Klux

-0.79

Dos

-0.78

 Norris

-0.67

AVG

-0.63

 Typhoon

-0.58

 Dame

-0.58

iper

-0.58

 subpoen

-0.57

Kru

-0.57

sts

-0.57

POSITIVE LOGITS

based

1.16

driven

0.99

laden

0.98

oriented

0.98

bearing

0.95

matter

0.94

seeking

0.93

of

0.92

heavy

0.92

rich

0.91

Activations Density 0.078%

phrases related to evidence or proof

negative phrases related to validity or evidence

No Comments

No Known Activations