INDEX

Explanations

simple and straightforward concepts

oai_token-act-pair · gpt-3.5-turbo

instances of the word "simple"

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

âĹ¼

-0.87

 reckoned

-0.78

 largeDownload

-0.73

raints

-0.71

 Experts

-0.70

ITNESS

-0.68

Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯

-0.67

hips

-0.67

 Intellectual

-0.63

 Nanto

-0.63

POSITIVE LOGITS

tons

1.18

 simple

1.04

wallet

0.95

simple

0.94

 straightforward

0.94

minded

0.92

json

0.89

Simple

0.86

 syrup

0.81

 arithmetic

0.78

Activations Density 0.017%

simple and straightforward concepts

instances of the word "simple"

No Comments

No Known Activations