INDEX

Explanations

words related to replacement or replication

oai_token-act-pair · gpt-3.5-turbo

references to replicas and replacement items

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

pants

-0.77

nuts

-0.76

nut

-0.70

WARE

-0.69

++++++++++++++++

-0.68

uay

-0.67

hog

-0.67

ãĥ£

-0.67

mph

-0.65

 DRAG

-0.65

POSITIVE LOGITS

icas

1.20

acements

1.18

icates

1.17

acer

1.12

icator

1.11

icated

1.08

acement

1.05

icating

1.04

ications

1.02

acers

1.00

Activations Density 0.011%

words related to replacement or replication

references to replicas and replacement items

No Comments

No Known Activations