INDEX

Explanations

references to grievances and related procedures, as well as details about errors and returns in a transactional context

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_33/width_131k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.33.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 казалось

-0.37

knew

-0.37

 Pourtant

-0.35

Xna

-0.35

迫

-0.34

 denn

-0.34

 aslında

-0.34

而是

-0.34

必要

-0.33

又有

-0.33

POSITIVE LOGITS

 must

0.83

 cannot

0.81

 CANNOT

0.75

 shall

0.72

 MUST

0.67

 apapun

0.64

 ProtoMessage

0.63

must

0.63

 typelib

0.62

 darf

0.61

Activations Density 0.095%

references to grievances and related procedures, as well as details about errors and returns in a transactional context

No Comments

No Known Activations