INDEX

Explanations

references to entities ending in "-ners," possibly indicating names or titles

oai_token-act-pair · gpt-3.5-turbo

references to collaborations or partnerships

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

 restraining

-0.63

 vacuum

-0.61

 instance

-0.61

URA

-0.58

 HuffPost

-0.58

 abortions

-0.57

 horrors

-0.57

UTE

-0.57

 background

-0.56

Alz

-0.56

POSITIVE LOGITS

ners

1.65

linger

1.22

hip

1.18

ingen

1.12

pace

1.09

ettings

1.08

peed

1.07

ner

1.04

poon

1.02

hots

0.99

Activations Density 0.004%

references to entities ending in "-ners," possibly indicating names or titles

references to collaborations or partnerships

No Comments

No Known Activations