INDEX

Explanations

instances where the reader is directly addressed or instructed to take a particular action

oai_token-act-pair · gpt-3.5-turbo

the word "you" in various contexts

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

ipal

-0.82

Lago

-0.69

emon

-0.66

weight

-0.64

acular

-0.62

 Kemp

-0.62

ortality

-0.61

ģ«

-0.61

efe

-0.60

 Parameters

-0.59

POSITIVE LOGITS

're

1.23

 guys

1.18

tub

1.03

'll

1.00

 mileage

0.95

've

0.94

RS

0.84

'd

0.84

 yourselves

0.82

tu

0.82

Activations Density 0.217%

instances where the reader is directly addressed or instructed to take a particular action

the word "you" in various contexts

No Comments

No Known Activations