INDEX

Explanations

names of individuals or possibly locations

oai_token-act-pair · gpt-3.5-turbo

names of individuals and possibly references to affiliations or roles they hold

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

 USPS

-0.68

theless

-0.68

 supers

-0.66

 Dire

-0.66

 FedEx

-0.65

 horizont

-0.64

 Hyundai

-0.64

CLS

-0.63

 Coastal

-0.63

ModLoader

-0.62

POSITIVE LOGITS

aney

0.91

beck

0.90

acci

0.90

rower

0.87

rage

0.87

riott

0.86

zinski

0.86

mberg

0.85

oval

0.85

isner

0.85

Activations Density 0.076%

names of individuals or possibly locations

names of individuals and possibly references to affiliations or roles they hold

No Comments

No Known Activations