INDEX

Explanations

individuals and their affiliations or roles

oai_token-act-pair · gpt-3.5-turbo

instances of the word "who" in relation to individuals and their descriptions or roles

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

 economical

-0.67

 misogyn

-0.67

Georg

-0.64

 arbitrary

-0.64

urg

-0.63

destruct

-0.63

ÂłÂłÂłÂł

-0.62

 logical

-0.62

Failure

-0.61

Nut

-0.59

POSITIVE LOGITS

 oversaw

1.20

 oversees

1.16

 participated

1.07

 attended

1.06

 owns

1.05

 attends

1.02

 specializes

1.02

 specialize

1.01

 authored

0.96

 chaired

0.95

Activations Density 0.100%

individuals and their affiliations or roles

instances of the word "who" in relation to individuals and their descriptions or roles

No Comments

No Known Activations