INDEX

Explanations

mentions of specific forms of online communication like blog posts, Facebook posts, site posts, subreddit posts, speeches, letters, memos, telephonic briefings, tweets, and press releases

oai_token-act-pair · gpt-3.5-turbo

references to various types of written communications and announcements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

oiler

-0.64

 idols

-0.62

Mods

-0.56

 extremes

-0.55

 obstacles

-0.54

 corpses

-0.54

illions

-0.54

 menace

-0.53

ãĥİ

-0.53

 projectiles

-0.51

POSITIVE LOGITS

 released

0.94

 obtained

0.90

 accompanying

0.88

 published

0.85

 titled

0.83

 Thursday

0.80

 Tuesday

0.79

 Wednesday

0.78

 aired

0.78

 posted

0.78

Activations Density 0.128%

mentions of specific forms of online communication like blog posts, Facebook posts, site posts, subreddit posts, speeches, letters, memos, telephonic briefings, tweets, and press releases

references to various types of written communications and announcements

No Comments

No Known Activations