INDEX

Explanations

adjectives or adverbs denoting the highest degree or superlative form of an attribute

oai_token-act-pair · gpt-3.5-turbo

phrases that emphasize 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 @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

-0.71

heid

-0.67

 Nero

-0.60

 Fist

-0.60

rompt

-0.59

instead

-0.59

Tam

-0.57

onut

-0.56

 Reloaded

-0.56

 Creed

-0.56

POSITIVE LOGITS

 importantly

0.99

likely

0.98

 likely

0.88

 prominently

0.82

mos

0.81

 notably

0.78

 important

0.77

afa

0.75

 obvious

0.74

OWER

0.73

Activations Density 0.064%

adjectives or adverbs denoting the highest degree or superlative form of an attribute

phrases that emphasize the word "most" in various contexts

No Comments

No Known Activations