INDEX

Explanations

superlatives such as "most" and descriptions of prominence or importance

oai_token-act-pair · gpt-3.5-turbo

the word "most" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

pload

-0.75

 Travels

-0.69

rompt

-0.65

instead

-0.63

heid

-0.61

arium

-0.61

 Handling

-0.61

 Races

-0.60

 CARD

-0.58

ovie

-0.58

POSITIVE LOGITS

 importantly

0.94

 important

0.93

likely

0.86

 notable

0.86

 prominent

0.80

 prominently

0.79

 likely

0.79

worst

0.78

egu

0.77

 obvious

0.77

Activations Density 0.086%

superlatives such as "most" and descriptions of prominence or importance

the word "most" in various contexts

No Comments

No Known Activations