INDEX

Explanations

attends to the list separator "pipe" symbol from subsequent tokens, typically found in sequences of items or lists.

oai_attention-head · gpt-4o Triggered by @mlfinancebook

New Auto-Interp

Configuration

jbloom/Gemma-2b-Residual-Stream-SAEs/gemma_2b_blocks.6.hook_resid_post_16384_anthropic_fast_lr

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

HuggingFaceFW/fineweb

Features

16,384

Data Type

torch.float32

Hook Point

blocks.6.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

HuggingFaceFW/fineweb

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 unlaw

-1.73

 despotism

-1.65

 belliger

-1.59

 disgra

-1.54

 ruinous

-1.49

 unspeak

-1.46

 impractica

-1.46

 Fascism

-1.39

 massacres

-1.38

 odious

-1.37

POSITIVE LOGITS

<bos>

17.30

GEBURTSDATUM

2.77

expandindo

2.76

 betweenstory

2.73

 Administrativna

2.62

تقاوى

2.61

Autoritní

2.58

 Italijani

2.44

 Италијани

2.44

 Мексичка

2.39

Activations Density 0.175%

attends to the list separator "pipe" symbol from subsequent tokens, typically found in sequences of items or lists.

No Comments

No Known Activations