INDEX

Explanations

phrases indicating a sequence of information or priority

oai_token-act-pair · gpt-3.5-turbo

references to visibility or being observed

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

IUM

-0.75

Bir

-0.71

Gameplay

-0.66

lys

-0.62

Plot

-0.62

ingen

-0.60

urat

-0.60

ople

-0.59

MEN

-0.59

fell

-0.58

POSITIVE LOGITS

imester

0.77

 princ

0.75

 elig

0.73

yip

0.73

fm

0.69

cation

0.66

 bidder

0.64

 millisec

0.63

eve

0.62

 agre

0.60

Activations Density 0.118%

phrases indicating a sequence of information or priority

references to visibility or being observed

No Comments

No Known Activations