INDEX

Explanations

references to the word "nerd" or similar variations

oai_token-act-pair · gpt-3.5-turbo

references to individuals named "Ner" or "Ud" as well as variations of "nerd"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

achusetts

-0.78

croft

-0.75

Dispatch

-0.74

 Hopkins

-0.74

auga

-0.73

 jurisdictions

-0.73

 Kasich

-0.71

ct

-0.71

 sweep

-0.71

 Phillips

-0.70

POSITIVE LOGITS

Ner

4.19

ner

2.65

 Nerd

1.51

 nerds

1.33

Ud

1.33

 nerd

1.30

Fey

1.21

Pau

1.15

Nou

1.06

 darts

1.04

Activations Density 0.033%

references to the word "nerd" or similar variations

references to individuals named "Ner" or "Ud" as well as variations of "nerd"

No Comments

No Known Activations