INDEX

Explanations

positive or negative evaluations of different aspects within a context

oai_token-act-pair · gpt-3.5-turbo

phrases discussing the various aspects or elements of a situation or experience

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

 actionGroup

-0.72

flush

-0.65

 sidx

-0.64

 throats

-0.61

 teasp

-0.59

 prefer

-0.59

STATE

-0.59

ignt

-0.58

 fever

-0.57

perate

-0.57

POSITIVE LOGITS

 surprises

0.68

 happens

0.67

SPONSORED

0.66

 happened

0.64

uary

0.63

ï¸ı

0.62

Merit

0.62

 thing

0.62

 Mara

0.62

 Flavoring

0.61

Activations Density 0.109%

positive or negative evaluations of different aspects within a context

phrases discussing the various aspects or elements of a situation or experience

No Comments

No Known Activations