INDEX

Explanations

phrases related to federal agencies and their actions or requirements

oai_token-act-pair · gpt-3.5-turbo

references to press releases or news articles, often indicated by the term "AP."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Lauder

-0.75

ãĥĥãĥī

-0.71

 cigar

-0.69

ãĤ©

-0.68

 Luthor

-0.68

 cigars

-0.66

 Monstrous

-0.65

 Ludwig

-0.64

 Brand

-0.62

 Kern

-0.62

POSITIVE LOGITS

PLIC

1.18

TOP

1.11

olitan

1.09

PLE

1.07

oleon

1.02

PLIED

1.01

rison

1.01

earance

1.00

etition

0.99

ACK

0.99

Activations Density 0.009%

phrases related to federal agencies and their actions or requirements

references to press releases or news articles, often indicated by the term "AP."

No Comments

No Known Activations