INDEX

Explanations

questions asking for or providing information or knowledge

oai_token-act-pair · gpt-3.5-turbo

inquiries and prompts related to user engagement or knowledge

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

forth

-0.74

Leaks

-0.61

ENE

-0.60

orders

-0.58

 advise

-0.55

âĸĪâĸĪâĸĪâĸĪ

-0.54

 Canaver

-0.54

ãĥ¢

-0.53

 intim

-0.52

eters

-0.52

POSITIVE LOGITS

tu

0.75

baugh

0.74

 Favorite

0.71

bp

0.63

iked

0.61

culosis

0.58

avascript

0.57

 browser

0.57

Want

0.56

rawdownloadcloneembedreportprint

0.56

Activations Density 0.097%

questions asking for or providing information or knowledge

inquiries and prompts related to user engagement or knowledge

No Comments

No Known Activations