INDEX

Explanations

past events or locations where specific actions took place

oai_token-act-pair · gpt-3.5-turbo

references to specific locations and events

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

 Malone

-0.70

CBI

-0.66

abee

-0.64

gdala

-0.63

icism

-0.62

oner

-0.60

iasco

-0.59

 Changes

-0.59

illian

-0.58

 Cancel

-0.58

POSITIVE LOGITS

 congreg

0.85

ezvous

0.80

 resides

0.79

 reside

0.79

 resided

0.73

astery

0.72

imum

0.72

Assembly

0.70

 safely

0.69

 holiest

0.67

Activations Density 0.390%

past events or locations where specific actions took place

references to specific locations and events

No Comments

No Known Activations