INDEX

Explanations

common names (abbreviation CN) with a high level of similarity

oai_token-act-pair · gpt-3.5-turbo

abbreviations and acronyms related to organizations and entities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

adjusted

-0.72

dose

-0.72

à©

-0.69

à¨

-0.68

 Loll

-0.66

killers

-0.65

 Vaughan

-0.64

Redditor

-0.63

started

-0.63

maker

-0.63

POSITIVE LOGITS

CN

1.03

CN

0.97

avez

0.85

alogy

0.81

uggets

0.80

SS

0.80

emonic

0.79

FK

0.78

CHR

0.74

vier

0.72

Activations Density 0.019%

common names (abbreviation CN) with a high level of similarity

abbreviations and acronyms related to organizations and entities

No Comments

No Known Activations