INDEX

Explanations

adjectives or descriptions of characteristics that are particularly positive

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate something is "best known" or notable for a specific reason

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Manson

-0.64

kers

-0.62

idon

-0.62

ker

-0.61

 Emer

-0.61

 Reloaded

-0.61

 probing

-0.61

chy

-0.60

 procession

-0.59

Dru

-0.58

POSITIVE LOGITS

seller

1.14

iaries

1.11

iary

1.09

ows

1.08

sell

1.07

ow

1.03

owing

1.02

 suited

0.99

ower

0.99

imates

0.86

Activations Density 0.054%

adjectives or descriptions of characteristics that are particularly positive

phrases that indicate something is "best known" or notable for a specific reason

No Comments

No Known Activations