INDEX

Explanations

phrases that indicate curiosity or questioning

oai_token-act-pair · gpt-3.5-turbo

expressions of curiosity or questioning

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

idelines

-0.64

 Guidelines

-0.62

IDs

-0.61

HF

-0.61

Low

-0.60

Aff

-0.60

BA

-0.60

stra

-0.59

ICO

-0.58

 Locke

-0.57

POSITIVE LOGITS

 wonder

4.09

 wonders

2.45

 marvel

1.94

 wondered

1.87

 wondering

1.73

 Wonders

1.51

 doubt

1.39

 amaz

1.38

 aston

1.33

 pity

1.33

Activations Density 0.011%

phrases that indicate curiosity or questioning

expressions of curiosity or questioning

No Comments

No Known Activations