INDEX

Explanations

superlatives, particularly identifying items that are the most significant or extreme within a category

oai_token-act-pair · gpt-3.5-turbo

the word "most" in various contexts that indicate a superlative or the highest degree of something

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 6-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.6.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.6.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

 Films

-0.69

heid

-0.68

instead

-0.66

 Fist

-0.59

rompt

-0.59

Tam

-0.58

 Creed

-0.58

 Chips

-0.58

 Palace

-0.57

 CARD

-0.57

POSITIVE LOGITS

 importantly

1.08

likely

0.95

 likely

0.92

 prominently

0.84

 important

0.83

 notably

0.83

 assured

0.81

 noticeable

0.78

reet

0.77

 obvious

0.76

Activations Density 0.053%

superlatives, particularly identifying items that are the most significant or extreme within a category

the word "most" in various contexts that indicate a superlative or the highest degree of something

No Comments

No Known Activations