INDEX

Explanations

commands or functions related to execution processes and task management in programming contexts

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

run variations

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_32/width_131k/average_l0_88

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.32.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

 disambiguazione

-0.65

addPreferredGap

-0.52

 CreateTagHelper

-0.51

 informée

-0.51

 Numerade

-0.49

LabelTagHelper

-0.47

 ujednoznacz

-0.47

省市镇

-0.47

 myſelf

-0.46

енча

-0.46

POSITIVE LOGITS

run

0.78

RUN

0.73

Run

0.70

 running

0.69

run

0.69

Run

0.69

RUN

0.69

running

0.64

runs

0.64

 Running

0.63

Activations Density 0.009%

commands or functions related to execution processes and task management in programming contexts

run variations

No Comments

No Known Activations