INDEX

Explanations

The main thing this neuron does is find terms related to "Syn"onyms or "Syn"chronization

oai_token-act-pair · gpt-3.5-turbo

references to religious organizations or terms

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

ORED

-0.87

aneers

-0.74

sburgh

-0.71

-0.69

IVE

-0.69

IRD

-0.68

ERE

-0.67

uay

-0.67

lessness

-0.67

ffee

-0.65

POSITIVE LOGITS

agogue

1.26

Syn

1.21

opsis

1.15

Syn

1.04

agog

1.02

chron

1.01

syn

0.98

apse

0.98

onymous

0.94

emy

0.94

Activations Density 0.006%

The main thing this neuron does is find terms related to "Syn"onyms or "Syn"chronization

references to religious organizations or terms

No Comments

No Known Activations