INDEX

Explanations

organization or company names, especially if they contain an abbreviation

oai_token-act-pair · gpt-3.5-turbo

abbreviations or acronyms related to organizations and technical terms

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 weap

-0.72

 Jordanian

-0.65

perse

-0.61

nces

-0.61

dyl

-0.60

nesota

-0.60

ozyg

-0.59

Redditor

-0.57

 symp

-0.57

 piston

-0.57

POSITIVE LOGITS

0.94

),

0.92

VC

0.90

)'

0.90

FU

0.89

DAQ

0.89

TF

0.85

IC

0.83

OE

0.83

)—

0.81

Activations Density 0.071%

organization or company names, especially if they contain an abbreviation

abbreviations or acronyms related to organizations and technical terms

No Comments

No Known Activations