INDEX

Explanations

names with a specific structure: two short parts separated by a dash and potentially followed by other characters

oai_token-act-pair · gpt-3.5-turbo

references to specific locations or properties associated with notable individuals

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

 Extras

-0.71

LW

-0.67

 Grimes

-0.66

 Chambers

-0.64

imore

-0.64

 Flavoring

-0.64

kefeller

-0.63

 foregoing

-0.63

 Vessel

-0.61

DragonMagazine

-0.60

POSITIVE LOGITS

shaped

1.02

nom

0.96

sized

0.90

sama

0.89

san

0.89

chan

0.89

cell

0.88

cert

0.87

inf

0.86

induced

0.86

Activations Density 0.135%

names with a specific structure: two short parts separated by a dash and potentially followed by other characters

references to specific locations or properties associated with notable individuals

No Comments

No Known Activations