INDEX

Explanations

numbers that are answers to questions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "answer."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ovie

-0.76

wana

-0.76

ammy

-0.71

zinski

-0.70

erker

-0.69

 sights

-0.68

ker

-0.67

 compuls

-0.67

 Nanto

-0.67

enic

-0.66

POSITIVE LOGITS

 answer

1.25

answer

1.19

ysis

1.15

swers

1.15

 answ

1.07

 Answer

1.03

 answering

1.00

 answers

0.98

 answered

0.96

Answer

0.96

Activations Density 0.017%

numbers that are answers to questions

instances of the word "answer."

No Comments

No Known Activations