INDEX

Explanations

proper names or abbreviations typically in the format of "BN" followed by a number, often related to institutions or organizations

oai_token-act-pair · gpt-3.5-turbo

occurrences of the abbreviation "BN" and related bibliographic references

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

rano

-0.96

flies

-0.69

door

-0.68

ammy

-0.68

houses

-0.64

Â¯

-0.64

itation

-0.64

inav

-0.63

arted

-0.63

come

-0.63

POSITIVE LOGITS

BN

0.90

KK

0.83

FW

0.79

uggets

0.71

RG

0.71

aylor

0.70

MQ

0.70

ECA

0.69

 Laden

0.67

iversary

0.67

Activations Density 0.007%

proper names or abbreviations typically in the format of "BN" followed by a number, often related to institutions or organizations

occurrences of the abbreviation "BN" and related bibliographic references

No Comments

No Known Activations