INDEX

Explanations

terms related to specific historical events or legal terms like "genocide"

oai_token-act-pair · gpt-3.5-turbo

terms related to genocide and associated discussions

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

 Bark

-0.88

 Salman

-0.81

 Emirates

-0.79

 Rash

-0.75

ASA

-0.69

Redd

-0.68

Eye

-0.67

 Nottingham

-0.67

 fare

-0.67

Jah

-0.66

POSITIVE LOGITS

toc

1.78

roc

1.56

 Phelps

1.32

wic

1.30

Rac

1.19

Coc

1.10

 genocide

1.07

perm

1.07

phi

1.06

coc

1.04

Activations Density 0.067%

terms related to specific historical events or legal terms like "genocide"

terms related to genocide and associated discussions

No Comments

No Known Activations