INDEX

Explanations

conjunctions, particularly the word "and."

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

processes and analysis

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

écri

-0.47

íslu

-0.47

 planète

-0.45

 république

-0.42

 dimas

-0.42

Còn

-0.41

 nôtre

-0.41

Achter

-0.41

 patate

-0.40

voerd

-0.40

POSITIVE LOGITS

OGND

0.68

 evaluation

0.67

حياته

0.65

 initiation

0.63

 assessment

0.60

tagHelperRunner

0.59

揄

0.59

 maintenance

0.59

 manipulation

0.59

 repair

0.57

Activations Density 0.097%

conjunctions, particularly the word "and."

processes and analysis

No Comments

No Known Activations