INDEX

Explanations

mentions of "Brit" or "Brits."

oai_token-act-pair · gpt-3.5-turbo

references to British individuals or entities

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

keeping

-0.64

handler

-0.63

ledged

-0.63

asso

-0.63

urized

-0.62

·

-0.62

handled

-0.62

 delegation

-0.61

hood

-0.61

 Minotaur

-0.61

POSITIVE LOGITS

 Brit

4.03

Brit

3.13

 Britons

2.04

British

1.64

 Britann

1.55

 Scot

1.54

 Swed

1.52

 British

1.52

 Britt

1.50

 Spani

1.46

Activations Density 0.026%

mentions of "Brit" or "Brits."

references to British individuals or entities

No Comments

No Known Activations