INDEX

Explanations

names or terms related to technical specifications or events

oai_token-act-pair · gpt-3.5-turbo

frequent suffixes or endings in words

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

 Carib

-0.81

 Buchanan

-0.76

 gram

-0.73

-0.70

 cycl

-0.68

CBO

-0.68

Tot

-0.67

cro

-0.67

LOT

-0.67

 Cycl

-0.67

POSITIVE LOGITS

en

1.67

EN

1.45

ens

1.31

enf

1.28

ename

1.23

ening

1.18

ened

1.17

ena

1.15

endon

1.13

eners

1.11

Activations Density 0.107%

names or terms related to technical specifications or events

frequent suffixes or endings in words

No Comments

No Known Activations