INDEX

Explanations

words related to seeking or searching for various things such as documents, answers, warmth, work, profits, etc

oai_token-act-pair · gpt-3.5-turbo

terms related to searching for information, documents, or assistance

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

 proud

-0.62

 Notting

-0.58

 snail

-0.56

 Count

-0.56

fox

-0.55

 Plague

-0.55

 frown

-0.55

 Boone

-0.55

 Typhoon

-0.54

 discredited

-0.54

POSITIVE LOGITS

 explanations

0.71

thood

0.70

 inspiration

0.69

 answers

0.69

 injection

0.69

 clarification

0.68

 urgently

0.68

 attention

0.67

olutions

0.67

 ASAP

0.67

Activations Density 0.453%

words related to seeking or searching for various things such as documents, answers, warmth, work, profits, etc

terms related to searching for information, documents, or assistance

No Comments

No Known Activations