INDEX

Explanations

the word "Well."

oai_token-act-pair · gpt-3.5-turbo

the word "Well" 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

illary

-0.75

âĹ¼

-0.75

İĭ

-0.67

Ã³n

-0.65

 flair

-0.65

MX

-0.62

adena

-0.62

hyde

-0.59

 vain

-0.59

 lightsaber

-0.59

POSITIVE LOGITS

esley

1.10

come

0.89

espie

0.86

ington

0.86

ustration

0.77

 Enough

0.75

ness

0.75

oyd

0.73

nesses

0.72

FTWARE

0.72

Activations Density 0.029%

the word "Well."

the word "Well" in various contexts

No Comments

No Known Activations