INDEX

Explanations

It is difficult to determine the specific pattern that neuron 4 is looking for based on the provided activations

oai_token-act-pair · gpt-3.5-turbo

abbreviations and acronyms related to organizations or concepts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

orer

-0.67

fusc

-0.65

Redditor

-0.60

artz

-0.59

umers

-0.59

Zero

-0.59

erb

-0.58

ussion

-0.57

ãĥ¼ãĥĨ

-0.56

role

-0.56

POSITIVE LOGITS

VILLE

1.33

 CITY

1.22

 COUNTY

1.17

SHARE

1.14

FIELD

1.11

WASHINGTON

1.07

 PARK

1.05

 EDITION

1.04

TIT

1.04

SPR

1.02

Activations Density 0.133%

It is difficult to determine the specific pattern that neuron 4 is looking for based on the provided activations

abbreviations and acronyms related to organizations or concepts

No Comments

No Known Activations