INDEX

Explanations

proper names, specifically the name "Alfred" with a high activation value

oai_token-act-pair · gpt-3.5-turbo

mentions of specific names, particularly "Alfred" and "Gustav"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

cript

-0.79

leading

-0.79

stakes

-0.77

former

-0.76

boarding

-0.75

BOOK

-0.75

ItemImage

-0.70

TBD

-0.68

cv

-0.68

leaders

-0.67

POSITIVE LOGITS

sson

1.04

 Alfred

1.02

 corrid

0.99

sburg

0.94

 obser

0.86

ulla

0.84

ibrary

0.82

 destro

0.80

anca

0.78

 acquaintance

0.77

Activations Density 0.003%

proper names, specifically the name "Alfred" with a high activation value

mentions of specific names, particularly "Alfred" and "Gustav"

No Comments

No Known Activations