INDEX

Explanations

phrases related to the use of objects or concepts

oai_token-act-pair · gpt-3.5-turbo

references to skills or usage of certain tools and concepts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

DragonMagazine

-0.71

ãĤ·ãĥ£

-0.63

 timeline

-0.61

 timelines

-0.61

bringing

-0.60

comings

-0.60

joining

-0.59

Completed

-0.59

ORK

-0.58

FontSize

-0.58

POSITIVE LOGITS

 responsibly

1.31

 recre

1.20

 inconsist

1.17

 illegally

1.14

 sparing

1.13

 inappropriately

1.12

 indoors

1.12

 everyday

1.11

 religiously

1.11

 extensively

1.09

Activations Density 0.455%

phrases related to the use of objects or concepts

references to skills or usage of certain tools and concepts

No Comments

No Known Activations