INDEX

Explanations

phrases related to Q&A or questions and answers

oai_token-act-pair · gpt-3.5-turbo

references to question and answer formats in discussions or reports

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

wagen

-0.79

 hers

-0.70

fulness

-0.65

Pra

-0.63

 Painter

-0.63

 Crimson

-0.63

fitting

-0.61

ufact

-0.59

 delinqu

-0.59

Vol

-0.57

POSITIVE LOGITS

UE

1.32

WER

1.28

ubes

1.23

ues

1.18

addafi

1.11

atari

1.05

uran

1.03

ube

1.01

UI

1.01

wer

1.00

Activations Density 0.026%

phrases related to Q&A or questions and answers

references to question and answer formats in discussions or reports

No Comments

No Known Activations