INDEX

Explanations

keywords related to different types of information, such as audience reactions, rumors, hospitals bills, crowd excitement, fighting stats, package details, lyrics, coding, attitude, systems, turf, stuff, responses, losses, information, pie, explanations, games, holes

oai_token-act-pair · gpt-3.5-turbo

critical nouns that indicate important subjects or themes in a discussion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

ielding

-0.68

REE

-0.65

 Residents

-0.63

 ILCS

-0.62

azard

-0.62

SEE

-0.60

 Â§Â§

-0.59

 Comprehensive

-0.59

MORE

-0.58

 Local

-0.58

POSITIVE LOGITS

 sucks

1.14

ain

1.07

 doesnt

0.99

 wasn

0.97

guy

0.96

isn

0.95

 smelled

0.93

 doesn

0.92

 didnt

0.92

 hasn

0.91

Activations Density 0.497%

keywords related to different types of information, such as audience reactions, rumors, hospitals bills, crowd excitement, fighting stats, package details, lyrics, coding, attitude, systems, turf, stuff, responses, losses, information, pie, explanations, games, holes

critical nouns that indicate important subjects or themes in a discussion

No Comments

No Known Activations