INDEX

Explanations

names of directors

oai_token-act-pair · gpt-3.5-turbo

occurrences of directed or produced content along with attributions to individuals or entities

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

hello

-0.62

ratom

-0.62

 toast

-0.62

 status

-0.60

 inequ

-0.60

 situation

-0.59

sit

-0.58

omin

-0.57

cape

-0.57

mod

-0.56

POSITIVE LOGITS

lain

0.78

 Architects

0.72

berus

0.71

 consortium

0.71

 Ariel

0.71

 Arrows

0.67

Â¨

0.67

 onwards

0.67

ortium

0.67

QC

0.67

Activations Density 0.948%

names of directors

occurrences of directed or produced content along with attributions to individuals or entities

No Comments

No Known Activations