INDEX

Explanations

specific character sequences and unique identifiers related to technical or programming contexts

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

death throes

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

the neuron lights up on content-word stems (meaningful word roots/long lexical morphemes) — i.e., parts of substantive words rather than short function words.

oai_token-act-pair · gpt-5-mini Triggered by @vetterc0

New Auto-Interp

Top Features by Cosine Similarity

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

Configuration

google/gemma-scope-9b-it-res/layer_9/width_131k/average_l0_121

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.9.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

 Raub

-0.75

 Erbe

-0.72

 Erben

-0.70

 Strahl

-0.65

 Massen

-0.63

 Insel

-0.63

 Herren

-0.63

 Kriege

-0.63

 Herzen

-0.62

 Mangel

-0.62

POSITIVE LOGITS

 فريبيس

0.71

friger

0.64

dirond

0.60

iſten

0.60

unicip

0.57

äiv

0.53

ésult

0.52

erapeu

0.52

ouſ

0.52

 săptăm

0.52

Activations Density 3.972%

specific character sequences and unique identifiers related to technical or programming contexts

death throes

the neuron lights up on content-word stems (meaningful word roots/long lexical morphemes) — i.e., parts of substantive words rather than short function words.

No Comments

No Known Activations

specific character sequences and unique identifiers related to technical or programming contexts

death throes

the neuron lights up on content-word stems (meaningful word roots/long lexical morphemes) — i.e., parts of substantive words rather than short function words.

No Comments

No Known Activations